Allow keys to be mapped to sequences of commands

[Omnikar]

This allows the user to use keymappings like this in config.toml in order to bind to a sequence of commands:

"a" = ["open_below", "normal_mode"]

[cessen]

One design question I have (that I'm not sure we actually need to resolve in this PR) is how things like this should interact with the undo system.

Specifically, if I bind a sequence of edit commands to a key, I think I would probably expect that to get pushed onto the undo stack as a single item. The same applies to e.g. custom commands written in a plugin or in a future possibly-lisp-based config file. (Edit: and also macros. And there may be further things that just aren't coming to mind at the moment.)

Again, we may not actually need to resolve this question in this PR. Commands sequences are only intended as a stop-gap solution anyway, so I think it's okay if there's a little jank to them. But I wanted to raise the question anyway since it's at least related.

[Omnikar]

@cessen Yeah, I agree that it should be added to undo history as a single entry. But how would I go about doing that?

[cessen]

@cessen Yeah, I agree that it should be added to undo history as a single entry. But how would I go about doing that?

Since this is a stop-gap solution, I don't know if it's actually important to figure out before merging this PR. I mainly mentioned it just to plant the thought.

[Omnikar]

Okay, so other than that, I think there's only this left before this PR is ready?

[Omnikar]

If we were to address the undo behaviour, would it work to have Document store a HashMap<ViewId, bool> of whether a command sequence is active, and have methods which take a ViewId — one for beginning sequences which just sets inserts true into the hashmap for the view id, and another for ending sequences that sets the hashmap entry to false and then calls append_changes_to_history — and then change append_changes_to_history to only actually do anything if the bool corresponding to the view id is false?

[archseer]

I would skip the undo behavior right now but this should be doable by either composing into a previous transaction in history, or accumulating changes before calling append_changes_to_history in a similar vein to insert mode.

[archseer]

Can you add an usage example to book/remapping.md?