Canonicalize paths before stripping current dir as prefix

[jazzfool]

Fixes #6182

[pascalkuthe]

This is not the right fix.

Yes this does fix the very specific instance you ran into but it doesn't adress the root cause of the problem. Right now you only canonicalize the paths if they exist. (also you crash the editor if CWD doesn't exist, we do that elsewhere too but we really shouldn't do that). If clangd had sent a message for a file that is not yet saved to disk (but exists as a buffer in helix) the same problem would occur. Furthermore there also many other places where we canonicalize paths in helix where having the normal canonicalization/normalization function being unreliable is a problem too and can lead to similar bugs.

As I described here: #6182 (comment) you need to change get_normalized_path to canonicalize the first parent that exists (and normalize the rest with the existing code). You can then use that function to normalize cwd (and the argument) in get_relative_path

[pascalkuthe]

This looks great now, thanks for sticking with it! one minor nit

[pascalkuthe]

Suggested change

	.find(|path| path.exists())
	.and_then(|path| Some((path, dunce::canonicalize(path).ok()?)))
	.map(|(base, canonicalized)| {
	(
	canonicalized,
	path.strip_prefix(base).unwrap(/* base is an ancestor of path */).into(),
	)
	})
	.find_map(|base| {
	let canonicalized_base = dunce::canonicalize(base).ok()?;
	Some((canonicalized_base, path.strip_prefix(base)))
	})

just because a path exists doesn't mean cannonicalization necessary successed (Sadly). Using find_map ensures we only stoip once canonicalization succedes. It's also a bit cleaner.

[pascalkuthe]

(didn't mean to approve yet until you accept my suggestion 😅 )

Test suite also needs to ne updated in a couple places (the previous paths assumed that we wouldn't cannonicalize)

[pascalkuthe]

LGTM now

[the-mikedavis]

Ah whoops I didn't see the conflict with #6156. Could you rebase this?