* The main interface is the `FT` command.
  `FT` was an undocumented Video TECO command for etags/ctags lookups.
  I don't want to exactly copy its interface, though.
* `FT` allows looking up symbol names,
  definitions and references.
* ctags will be supported via ctags-lsp.
  LSP support is more powerful though and works without
  regenerating TAGS files all the time.
  The LSP will also allow you to customize auto-completions
  using SciTECO itself (i.e. by writing a language server
  in SciTECO).
* For multiple results, `FT$` can be used to cycle through
  results - this should mimic repeated `S$` or `N$`.
* `:FT...$` does a fuzzy search. IMHO it's not important
  to return a status integer instead. `FT` will only
  really be used in interactive mode.
* Document synchronization is supported via hooks from
  ring.c and via Scintilla notifications.
* Currently, the LSP communication is based on blocking
  GIOChannels. This means that a misbehaving hanging
  server could "lock up" the entire editor (FIXME).
  Only on ncurses you can always kill the subprocess by
  pressing CTRL+C.
  We need helper functions in spawn.c to read and write
  with interruptions.
* The textDocument/didChange notification transmits
  not only all edits, but all files' contents as well
  during initial synchronization.
  Therefore it is optimized to write and JSON-escape
  data without copying them around in memory and without
  destroying the buffer gap.
* Use $SCITECO_LSP to configure the language server.
  You can also use `tee` to capture stdin and stdout.
  Perhaps $SCITECO_LSP should be saved in .teco_session,
  so you can change it between projects?
* $SCITECO_LSP_ROOT is used to point to the project's
  root directory. session.tes will set it up automatically.

Previously almost all glyph-to-byte offset conversions consulted
Scintilla's line index and counted characters on the resulting line.
For instance a simple expression like `.+1J` would scan the same line
twice completely, which would be very slow on pathologically long lines.
Even insertions did that due to having to update the ^Y ranges.
If you repeat such an operation over all characters as in `<.+1:J;>`
you would have complexity O(n^2) for n = line length.
Only commands with an explicit relative nature like `C` and `A` would
use teco_view_glyph2bytes_relative() which scans beginning at dot
as long as the relative movement is less than 1024 glyphs.

Wit the new heuristics almost all glyph-to-byte and byte-to-glyph
conversions can make use of that optimization.
This requires that dot must at all times be known in glyphs as well -
the byte position is managed by Scintilla (SCI_GETCURRENTPOS).
We therefore introduced teco_current_doc_set_dot() and
teco_current_doc_get_dot() to update dot in the current buffer or
Q-Register -- it cannot be stored along with the view since
Q-Registers share a single view.
A number of auxiliary functions have been introduced for
converting relative to a known (glyphs,bytes) offset pair
and for converting absolute and relative positions with regard
to the current doc and SCI_GETCURRENTPOS position.
Of course this is error-prone since the glyph and dot positions
are interdependant - they must always be kept in sync.

With these new optimizations even pathologically long lines can
(usually) be managed even in UTF-8 documents.
It does not address slow-downs in Scintilla's line layout, yet.
grosciteco.tes for instance runs twice as fast now.

* When passing a struct that should not be modified, I usually use a const pointer.
* Strings however are small 2-word objects and they are often now already passed via separate
  `gchar*` and gsize parameters. So it is consistent to pass teco_string_t by value as well.
  A teco_string_t will usually fit into registers just like a pointer.
* It's now obvious which function just _uses_ and which function _modifies_ a string.
  There is also no chance to pass a NULL pointer to those functions.

* Requiring state callbacks by generating their names (e.g. NAME##_input) has several disadvantages:
  * The callback is not explicitly referenced when the state is defined.
    So an unintroduced reader will see some static function, which is nowhere referenced and still
    doesn't cause "unused" warnings.
  * You cannot choose the name of function that implements the callback freely.
  * In "substates" you need to generate a callback function if you want to provide a default.
    You also need to provide dummy wrapper functions whenever you want to reuse some existing
    function as the implementation.
* Instead, we are now using static assertions to check whether certain callbacks have been
  implemented.
  Unfortunately, this does not work on all compilers. In particular GCC won't consider
  references to state objects fully constant (even though they are) and does not allow
  them in _Static_assert (G_STATIC_ASSERT). This could only be made to work in newer GCC
  with -std=c2x or -std=gnu23 in combination with constexpr.
  It does work on Clang, though.
  So I introduced TECO_ASSERT_SAFE() which also passes if the expression is *not* constant.
  These static assertions are not crucial - they do not check anything that can differ between
  systems. So we can always rely on the checks performed by FreeBSD CI for instance.
  Also, you will of course quickly notice missing callbacks at runtime - with and without
  additional runtime assertions.
* All mandatory callbacks must still be explicitly initialized in the TECO_DEFINE_STATE calls.
* After getting rid of generated callback implementations, the TECO_DEFINE_STATE macros
  can finally be qualified with `static`.
* The TECO_DECLARE_STATE() macro has been removed. It no longer abstracts anything
  and cannot be used to declare static teco_state_t anyway.
  Also TECO_DEFINE_UNDO_CALL() also doesn't have a DECLARE counterpart.

* It turns out that `bool` (_Bool) in bitfields may cause
  padding to the next 32-bit word.
  This was only observed on MinGW.
  I am not entirely sure why, although the C standard does
  not guarantee much with regard to bitfield memory layout
  and there are 64-bit available due to passing anyway.
  Actually, they could also be layed out in a different order.
* I am now consistently using guint instead of `bool` in bitfields
  to prevent any potential surprises.
* The way that guint was aliased with bitfield structs
  for undoing teco_machine_main_t and teco_machine_qregspec_t flags
  was therefore insecure.
  It was not guaranteed that the __flags field really "captures"
  all of the bit field.
  Even with `guint v : 1` fields, this was not guaranteed.
  We would have required a static assertion for robustness.
  Alternatively, we could have declared a `gsize __flags` variable
  as well. This __should__ be safe since gsize should always be
  pointer sized and correspond to the platform's alignment.
  However, it's also not 100% guaranteed.
  Using classic ANSI C enums with bit operations to encode multiple
  fields and flags into a single integer also doesn't look very
  attractive.
* Instead, we now define scalar types with their own teco_undo_push()
  shortcuts for the bitfield structs.
  This is in one way simpler and much more robust, but on the other
  hand complicates access to the flag variables.
* It's a good question whether a `struct __attribute__((packed))` bitfield
  with guint fields would be a reliable replacement for flag enums, that
  are communicated with the "outside" (TECO) world.
  I am not going to risk it until GCC gives any guarantees, though.
  For the time being, bitfields are only used internally where
  the concrete memory layout (bit positions) is not crucial.
* This fixes the test suite and therefore probably CI and nightly
  builds on Windows.
* Test case: Rub out `@I//` or `@Xq` until before the `@`.
  The parser doesn't know that `@` is still set and allows
  all sorts of commands where `@` should be forbidden.
* It's unknown how long this has been broken on Windows - quite
  possibly since v2.0.

* Test case: @EQa// @?/EX/ -- Rubout should return you to the Q-Register view.
* The test suite has been extended.
  Unfortunately we cannot currently directly check whether we're editing a Q-Register.
  But we add a magic number of characters to the Q-Register, that we can check afterwards.

* Curses allows scrolling with the scroll wheel at least
  if mouse support is enabled via ED flags.
  Gtk always supported that.
* Allow clicking on popup entries to fully autocomplete them.
  Since this behavior - just like auto completions - is parser state-dependant,
  I introduced a new state method (insert_completion_cb).
  All the implementations are currently in cmdline.c since there is some overlap
  with the process_edit_cmd_cb implementations.
* Fixed pressing undefined function keys while showing the popup.
  The popup area is no longer redrawn/replaced with the Scintilla view.
  Instead, continue to show the popup.