feat: add initial support for multi-stroke chord keybindings in emacs mode

[reubeno]

Adds basic support for adding multi-stroke (chord) keybindings, e.g.:

// Add chord bindings
// Ctrl+X Ctrl+C: Quit
keybindings.add_sequence_binding(
    &[ctrl('x'), ctrl('c')],
    ReedlineEvent::CtrlD,
);

Changes to public API:

• (Breaking change) Remove the existing pub field of Keybindings so we can unify the implementation between single-keystroke bindings and multi-keystroke bindings.

  I did some searching on GitHub and couldn't immediately find specific cases that would be broken by this -- but of course, that doesn't prove it won't break a client. (Most of the search hits that came up seemed to be vendored/forked copies of reedline or nushell, or similar field names in different types.)

• Adds Keybindings::add_sequence_binding() and Keybindings::remove_sequence_binding() to bind (or unbind) a sequence of 1 or more KeyCombinations to a ReedlineEvent; this supports both single-keystroke bindings and multi-keystroke chord-style bindings.

• Retains Keybinding::add_binding() and Keybinding::remove_binding() for backwards compatibility, replacing their bodies with simple wrapping calls to the new sequence-enlightened calls.

General approach:

• Do not allow a registered keystroke to be a strict prefix of another registered keystroke.
• Support arbitrarily long chords.
• Use a trie to store key bindings for efficient matching.
• Unify handling of single-keystroke keybindings with multiple-keystroke keybindings (do not special-case one vs. the other).
• Focus on emacs mode first; do not worry about aligning with existing multi-keystroke processing already present in vi mode for handling commands.

I'm looking for feedback on the general approach as well as the implementation before getting any deeper into this. For reference, I've demonstrated how this support could be used in nushell via this commit in my nushell fork. There's also a new example included in this draft PR to demonstrate and test.

(Aside: This is my first feature-level contribution. I'm eager to get as much feedback as I can!)

[fdncred]

Thanks for taking the time to put this up. I added a comment to your nushell commit above. I think the code seems plausible but I'm not a big fan of the configuration aspect of it.

[kronberger-droid]

Nice foundation.
Heads up that @benvansleen's #1016 is solving the same chord problem from the other end and I think the trie you put in Keybindings is the cleaner base of the two, so it'd be great to join forces.

A couple of things #1016 already handles that this is missing:
a timeout flush (needed for jj-in-insert, where a lone j has to self-insert after a pause), and the case where a key is both a binding and a prefix (j→Down and jj→Esc, which the trie currently disallows).
Could the two be combined... your trie as storage, his resolver + timeout on top as one shared piece the modes embed?

Up for coordinating?