Factored out execute function for the main loop

[tailhook]

Most commands are factored out to the function, except:

1. Commands that need extra input
2. Suspend, which needs access to the terminal

This should be enough to implement ContextValidator::invoke.

Also this commit brings CompleteHint, SelfInsert, Insert commands
to the match over Cmd instead of keeping them as separate if
statements as latter doesn't seem to change their behavior.

This is needed to implement #453, and #293 (although, latter might need
input-accepting commands too or another approach).

---

This is somewhat super-minimal refactoring to move closer to our goal of invoking commands from ContextValidator::invoke or any other user hooks.

I think input_mode.is_emacs_mode() and a reference to kill_ring can also be incorporated into State object. So the only thing we need to propagate through the call chain is State. What do you think? (this can be a follow-up PR anyway)

[gwenn]

For #293 (more precisely, for composite command), we need to cumulate impacts of each action and update output accordingly only once.
But your PR is still useful.

[tailhook]

For #293 (more precisely, for composite command), we need to cumulate impacts of each action and update output accordingly only once.

Well, I think we can do the following:

1. Drop most move_cursor and refresh_* in the edit.rs
2. Refresh line by looking at growing changes
3. If line is not refreshed move cursor by comparing to old position

(sans few special commands, like Undo)
Then ValidateContext::invoke also would only do the "impact" and and line is refreshed afterwards.

Thoughts?

But your PR is still useful.

Okay. Let's merge it and proceed quickly. Because this is a kind of change which will conflict with every addition or even change or other command (including #452) and keeping track of which commands are added in the master and which are already there in this branch is time-consuming.

[tailhook]

Rebased on the new master. So do you accept this as a first step? As I've said rebasing this PR every time a big piece of error-prone work so I don't want to keep it in a branch for too long.

[gwenn]

Sorry, the conflict was introduced by your PR #452...

[tailhook]

Sorry, the conflict was introduced by your PR #452...

Sure, fixed that.

[gwenn]

Thanks.