fix(window): improve cursor screen position calculation for multibyte chars

[petobens]

Update get_cursor_screen_position to use display width for accurate column indexing and add error handling for screenpos failures.
This ensures correct positioning with wide or multibyte characters and prevents runtime errors.

Fixes #1851

[Saghen]

:h screenpos says that it should take a byte-index for the column so I don't think we should be using the string display width here. In #1874, eokoshi mentions that the window id from nvim.api.nvim_win_get_number(0) appears to be wrong, which would explain why it's complaining about the line not existing.

[petobens]

Indeed the fix suggested in #1874 seems to work and is cleaner.

[pedro757]

I'm using this and it's fine, doesn't break anything! waiting for the merge!

[Saghen]

I'd like to figure out why we need to use the win_getid rather than blindly accepting. Afaict, nvim.api.nvim_win_get_number(0) and vim.fn.win_getid() should return the same window id.

[soifou]

I'd like to figure out why we need to use the win_getid rather than blindly accepting

Because this bug has been introduced by upstream commit eeacd7bd71 and from what I can tell, this change made the parent window logic stricter. So we cannot rely on nvim.api.nvim_win_get_number anymore?

We can use also simply 0:

local pos = vim.fn.screenpos(0, cursor_line, cursor_column)

[Saghen]

Thank you all!