refactor: remove the crate bstr from the dependency

[jaudiger]

This PR simplifies a bit the brioche-runtime-utils codebase by removing the crate bstr to instead rely exclusively on methods from std. The main advantage to use the methods from the standard is there are infaillible which removes a bunch of indirection in the code that were related to error handling (which explains the removal of 100 lines).

We still need bstr but it's now a transitive dependency.

[jaudiger]

On main:

With this PR:

[kylewlacy]

I think it'll take some research on my part before I'm ready to merge this 😅

I looked at the docs for both std::ffi::OsStr and bstr on how paths and OS strings are converted to/from bytes, and I'm not clear if bstr and std are equivalent here or not. We don't support Windows (today) so that's probably not a factor, but I'm not sure how each handles e.g. a path with an invalid UTF-8 sequence. (I'm also not sure if this is a real concern today since I also don't know if we have any packages today that use non-UTF-8 paths anywhere....)

[kylewlacy]

LGTM

I'm honestly still not sure about how bstr's from_bytes vs. the std library's works, and I don't really feel that strongly about keeping / removing bstr honestly (it's a nice convenience but not for how we're using it here).

I'm good to get rid of it though. It seems pretty low-risk overall, given how slowly this repo evolves and how it really gets put through its paces when brioche-packages updates it

@jaudiger let me know if you want me to take over resolving conflicts or anything here

[jaudiger]

Thanks @kylewlacy, I manually resolved the merge conflicts, and set auto-merge.