map C-f in Emacs mode to complete hints iff at end of line

[joooooooooooooooooooooooooooooooooooosh]

This doesn't affect the behaviour of C-f if in the middle of a line, so I expect there'd be minimal unexpected impact.
This is just a QOL improvement that would be nice to have for C-f (even though the Right arrow key already does this), as popular programs like zsh auto-suggestions already have this behaviour for C-f and there is very little current support for conditional/overloading mappings through the API for rustyline.

[gwenn]

there is very little current support for conditional/overloading mappings through the API for rustyline.

https://github.com/kkawakam/rustyline/blob/master/examples/custom_key_bindings.rs#L52

[joooooooooooooooooooooooooooooooooooosh]

I'm aware that this is doable (if a bit messy) using ConditionalEventHandler, but in this case since C-f wouldn't do anything normally when at the end of a line, is there a particular reason not to add extra behavior here? Given the Right arrow key already does this it feels natural to have C-f also be able to expand hints at the end of a line.

[gwenn]

I'm aware that this is doable (if a bit messy) using ConditionalEventHandler, but in this case since C-f wouldn't do anything normally when at the end of a line, is there a particular reason not to add extra behavior here? Given the Right arrow key already does this it feels natural to have C-f also be able to expand hints at the end of a line.

The Right arrow key behaviour was changed before ConditionalEventHandler.
See #533 for a similar discussion.
Should we

• introduce a zsh / zle mode,
• diverge further from emacs mode (as implicitly defined by readline) when principle of least surprise applies,
• ...
  ?

[joooooooooooooooooooooooooooooooooooosh]

For both this PR and #533, I would still argue that the principle of least surprise applies more to having C-[npbf] match the behaviour of the relevant arrow keys within the same application.
I don't think making the behaviour of these keys diverge from one another simply because the behaviour of the arrow keys were changed before ConditionalEventHandler was added is a good idea.

If there are more differences between common zsh bindings and readline bindings, it would probably be worth creating a seperate zsh / zle mode, but either way I think it's worth trying to make arrow keys consistent with their C-[npfb] counterparts in both modes.