Improve IME and text input support.

[bullno1]

• Allow grabbing the entire text input buffer. At least it is useful for my IME which sends an entire composed word.
• Replace rather than append composition. It is what the IME is "cooking". At least on Linux, the IME handles text replacement and the application just have to render.
• Add an IME example. Only tested with https://en.wikipedia.org/wiki/UniKey_(software) on Linux but SDL should be portable.

There are a few things that may need further thoughts:

Should cf_draw_text accepts NULL as input and treat it as an empty string? CF allows initializing a string as NULL and manipulate it. So NULL is equivalent to an empty string.

Do we need a way to render text from utf32? The current example keeps 2 concurrent representations: utf8 and utf32. utf32 is for editing since each codepoint is a glyph. utf8 is for rendering since CF expects it but upon rendering, it is decoded into utf32 anyway. That said, I'm not sure how text editors do this. I only need short textboxes so it doesn't really matter that much.

What should cf_text_size return? Previously, I changed it to be inclusive of the entire text box. But to render a cursor on a multilline string, that is not enough. For example:

First line is long
Second | <- cursor is here.

Maybe a CF_TextInfo cf_text_info(const char* text, int num_chars) where CF_TextInfo can accommodate different types of info.

Supporting IME composition candidates (think auto complete for IME): https://wiki.libsdl.org/SDL3/SDL_HINT_IME_IMPLEMENTED_UI#remarks. It is apparently a thing but I have no knowledge of this nor do I need it soon.

[RandyGaul]

Looks very well written!

[RandyGaul]

Should cf_draw_text accepts NULL as input and treat it as an empty string?

Yes.

Do we need a way to render text from utf32?

Probably not.

What should cf_text_size return? Previously, I changed it to be inclusive of the entire text box.

Sounds correct to me.

Maybe a CF_TextInfo cf_text_info(const char* text, int num_chars) where CF_TextInfo can accommodate different types of info.

This is a good direction. Let the user have more optional info about the text glyphs, and we can just pull it out from the font and hand it back. I don't want to clutter CF with anything too much or too specific. We want to offload more custom or nuanced stuff to the user.