feat: remove peer relation and improve integration tests

[jedel1043]

Addresses the review comments about the charm and the library.

• Removes the peer relation data in favor of a single unified autofs mounts file.
• Renames everything to be a bit more consistent with the "filesystem" prefix.
• Reworks client to only allow a single filesystem provider.

Since the previous integrations tests were too tied to the implementation, this replaces them with decoupled tests that only assert that the client can access remote files from the servers.

[wolsen]

Looking good! A few inline comments please.

[NucciTheBoss]

Looking pretty darn good 🤩

Just some suggested revisions around type annotations and other implementation bits. One thing I noticed is that some functions/methods don't have an annotation if they return None. E.g. the function/method doesn't have a return statement. Generally I include the -> None since it quickly signifies that there's no special sauce we expect in return when reading source code.

Another thing is that I believe we possibly had an agreement that log, exception, and unit/application messages would follow the Go convention where it's all lowercase, and messages are not postfixed with a period. It's the style we follow in slurm_ops. I say I believe because I remember we talked about it in a PR for slurm_ops, but we didn't write it down anywhere 😅

Let me know if you have any questions!

[NucciTheBoss]

Also, wanted to mention in writing from our discussion yesterday that I agree that we should look at making a monorepo with the proxy-operators include so that we don't have to create a "meta" charm within the integration tests and can just test directly against the proxies.

[jedel1043]

Another thing is that I believe we possibly had an agreement that log, exception, and unit/application messages would follow the Go convention where it's all lowercase, and messages are not postfixed with a period. It's the style we follow in slurm_ops. I say I believe because I remember we talked about it in a PR for slurm_ops, but we didn't write it down anywhere 😅

I think it was on charmed-hpc/hpc-libs#13 (comment), but the justification for that was the nice display of errors if they're lowercase (Error: message instead of Error: Message). For log messages and exceptions I completely agree, but for status messages Juju doesn't append anything behind the message, so it's better style to begin those in uppercase.

[NucciTheBoss]

I think it was on charmed-hpc/hpc-libs#13 (comment), but the justification for that was the nice display of errors if they're lowercase (Error: message instead of Error: Message). For log messages and exceptions I completely agree, but for status messages Juju doesn't append anything behind the message, so it's better style to begin those in uppercase.

This is an agreeable path forward for me. We'll need to do a deeper UX drive as we get closer to the Plucky release, but we'll hold off for after midcycle to really deep dive on how the charms should look and feel.

[NucciTheBoss]

LGTM! Excellent work

All suggestions were addressed/implemented. Dismissing because even though all conversations were resolved, GitHub seems to be having a cache issue somewhere that is preventing me from rebasing and merging this PR.