feat(colors): optionally get the cursor color from the hl group at the cursor

[folke]

When cursor_color = "none", we try using the color at the cursor's hl group as the cursor color.
This is similar to how kitty does the smearing cursors, with cursor none, meaning it takes the color of what's at the current cell as the cursor color.

When getting the color, we first try treesitter captures if treesitter highlighting is enabled in the buffer and then fallback to extmarks. Regular treesitter extmarks are ephemeral, so they can't be retrieved with the get extmarks function.

Everything is cached, so should not impact performance.

I believe you tried something similar, as there was a non-used function called M.get_hl_group already :)

I used some metatable magic to prevent making changes in other places where the hl group lookup tables were used. Let me know if you'd rather have those places call the function directly.

Changes

Mainly changes to the color module, where hl groups are now cached and dynamically created when needed.

[folke]

We need to loop over all captures. The last one is the one we want, so no early break here.

[folke]

In the old code, for level=0, it used to be "none" and for level>0 it was _normal_bg.
I assume this was a bug? And you always want none here?

[folke]

That issue is now fixed.
Will also add support for other extmarks.

[folke]

Added support for both treesitter and extmarks

[folke]

converted to draft again. Need to optimize the code a bit first

[folke]

Changed it so that the color at the cursor only gets updated on CursorMoved.

Much better for performance

[sphamba]

Thanks a lot, it works great!
I'll merge this and add an entry in the Readme. I'll also remove the metatables and directly call the get_hl_group function, but thanks for keeping the scope of your changes to the color module :)