Gracefully handle errors in CLI

[MichaReiser]

Summary

Replace the remaining unwrap calls in the CLI with more gracefully propagating the errors.

Long term, my goal is to replace anyhow::Error with a more structured representation (ideally it's just another form of Diagnostic)
because these errors don't look as nice as they could. But this matches Ruff's current behavior

Test Plan

Examples:

ruff failed
  Cause: Provided current-directory path '../test/a.py' is not a directory.

ruff failed
  Cause: Failed to search the virtual environment directory for `site-packages`
  Cause: No such file or directory (os error 2)

[github-actions]

ruff-ecosystem results

Linter (stable)

✅ ecosystem check detected no linter changes.

Linter (preview)

✅ ecosystem check detected no linter changes.

[carljm]

Nit: s/unicode/UTF-8/.

We have no idea if it contains unicode or not, because unicode is just a set of numeric code points, it's not an encoding. It may well be valid Unicode encoded as UTF-32, for all we know. It's just not valid UTF8-encoded text.

[MichaReiser]

I'm not so sure if we're limited to UTF8 here. Path::to_string_lossly specifically mentioned unicode and not UTF8

[carljm]

It only handles UTF-8. (Well, on Windows it handles a weird Windows-specific variant of UTF-8, but close enough.) It's not possible to transparently and automatically handle all the various Unicode encodings; they are ambiguous with each other. You have to either assume or specify a particular encoding.

It's too bad that the rust docs are also careless with this wording. It's meaningless to say that a byte string "is Unicode" or "is not Unicode", unless by "Unicode" you mean (or assume) some specific encoding.

[MichaReiser]

Yeah, still not sure about this. It's true that you can only store one encoding, but you can decode between encodings. E.g. Rust converts the Windows UTF16 paths to UTF8. That means, it isn't about what unicode encoding paths use on the file system, for as long as Rust decodes the paths to UTF8 before hand.

Also, from the camino docs https://docs.rs/camino/latest/camino/

Either way. I think using Unicode here is "correct" enough.

[carljm]

I see. You (and probably the rust docs too) don't want to say UTF-8 (even though the actual decode that's failing is definitely a decode from UTF-8, or the windows "wide" variant of it) because the original path on disk on windows is UTF-16 (decoded by rust and re-encoded as "wide" UTF-8 in the OsStr). So we use "Unicode" as imprecise shorthand for "the typical Unicode encoding used on your OS, ie UTF-16 on Windows and UTF-8 everywhere else." I guess maybe that's reasonable, for lack of a more precise term.