Implement changing editing mode

[ima1zumi]

I implemented Readline’s emacs-editing-mode (C-e) and vi-editing-mode (M-C-j, M-C-m) to switch editing modes. The keymap was aligned with Readline's behavior. But M-C-m was used in the yamatanooroti tests, so I decided not to map it.

[st0012]

Since both mappings are currently used for other operations, this feels like a breaking change.
Do we know if people use the current mappings and did we receive request for the new behaviour?

[ima1zumi]

The original key mapping is different from Readline, so I believe few people use it.
On second thoughts, not many people, other than maintainers, want to change between Vi mode and Emacs mode using the default key bindings. I think I should stop binding them by default. Thank you for your comment!

[tompng]

Just for information, GNU Bash source code comment also says it's confusing.

  /* In Bash, the user can switch editing modes with "set -o [vi emacs]",
     so it is not necessary to allow C-M-j for context switching.  Turn
     off this occasionally confusing behaviour. */
  ...
  if (func == rl_vi_editing_mode)
    rl_unbind_key_in_map (CTRL('J'), emacs_meta_keymap);
  ...
  if (func == rl_vi_editing_mode)
    rl_unbind_key_in_map (CTRL('M'), emacs_meta_keymap);
  ...
  if (func == rl_emacs_editing_mode)
    rl_unbind_key_in_map (CTRL('E'), vi_movement_keymap);

[ima1zumi]

When key binding is set, key.char is passed as symbol to LineEditor#update.
I can't seem to test it properly with input_keys. If you have a better way to do this, please let me know.