stb_textedit: Add support for custom move word left/right handlers

[ocornut]

EDIT Minor edits applied since first posting.

Sorry to submit that right after you've done a bunch yesterday!

This patch allows the user to optionally define two functions to override ctrl-arrow style per-word movement, STB_TEXTEDIT_MOVEWORDLEFT(obj,i) and STB_TEXTEDIT_MOVEWORDRIGHT(obj,i)

STB_TEXTEDIT_IS_SPACE(ch) is still supported and will generate default handlers for those, matching the existing code.

The reasoning is to allow applications to:

• Support OS X style word-jumping (move-to-next-word moves to the end of the current word instead of moving you to the beginning of the next word) to have an app be more friendly to OS. Someone made various mods to ImGui to match OS X behaviour and that was the only change they wanted that had to be done on textedit-side.
• Perform other/custom tailored word-jumping, where different tokens in a programming language may lead to slightly different moving behaviour.

Minor notes:

• I called them MOVEWORDLEFT/MOVEWORDRIGHT to match the WORDLEFT-WORDRIGHT terminology expressed in the public key definitions, whereas some of the internal code used "previous" "next". Also note that the existing STB_TEXTEDIT_IS_SPACE with underscore should be called ``STB_TEXTEDIT_ISSPACE` to strictly be consistent with the other function names. It doesn't really matter, but that is why the new define don't have underscores.
• I put the "burden" of passing state->cursor-1 or state->cursor+1 to the main key handling switch in textedit. Depending what you feel is best, the switch could pass state->cursor to both those functions and leave it to the handlers to do -1/+1. It may actually make more sense like that! (just realizing this as I'm typing this message, I could amend the PR for that if you want).
• Coding style: Some functions near the affected code used _arg for function arguments instead of arg commonly used in the rest of the code. I haven't changed that to minimize the patch overhead but kept one argument name c also to reduce the patch size and make it easier for you to parse it. I'm not sure if there was a reasoning for the extra underscore or if it came accidentally with another patch. While you are there you might want to rename the very few occurences of _str to str to make the code more consistent.

Thanks!

[nothings]

I cleaned up the obvious initial-underscore identifiers, and I removed the forced -1/+1 in the keyhandler as you suggested so when you switch to this version you'll need to adjust your existing handlers to compensate.

Unfortunately I'm not set up to test textedit, so these changes compile but it's possible I broke something.