SFTP Sync Support

[OpenBagTwo]

Summary

Addresses #42 by implementing pure-Python syncing via SFTP and Paramiko

List of Changes

• Updates pre-commit
• Creates a new "sftp" extras that will install paramiko and its dependencies
• Refactors test_sync to separate out tests of the sync utilities into their own test module
• Instead of a crude "nuke everything and copy over the source," file-protocol syncing now only copies / deletes the individual files that have changed. This makes it an actual viable sync method, as per the documentation, as opposed to something that would be horribly inefficient if used in reality / outside of testing
• All protocols now generate useful dry-run sync reports
• Implements differential SFTP-based syncing via paramiko
• Tests that all three protocols implement and respect the delete flag (which, FWIW, isn't currently exposed through the CLI)
• Adds some documentation on syncing via SFTP

Tech Debt and Other Concerns

• The diffs do not actually compute checksums and thus are prone to false-negatives. But I don't believe rsync actually computes diffs any more accurately.
• The file and sftp syncs copy/paste a lot of code, but so far I haven't settled on a good / general abstraction. If/when I tackle #43 will probably be the time to refactor.

To Do

• [x] The diff tests will occasionally fail if the files aren't created within the precisely same millisecond. This is a nasty bug, since it doesn't occur every time (though it seems to happen more reliably on slower systems like the instances used for these actions :grin:), so one solution might be to @xfail these tests (and move this bullet into "Tech Debt")
• [x] Use the verbosity and timeout sync args, or lose them
• [x] Implement Windows test runner
• [x] Flesh out SFTP documentation and include an example workflow of how you would, say, sync your ROG Ally and your Steam Deck
• [x] There's a lot of untested code (coverage is sitting at 83% right now), and looking over the missings, I definitely think there's room for more tests

Validation Performed

• [x] All gaps in testing have been rationalized (see "To Do"), and all unit tests pass or are validly skipped across all three operating systems (x both versions of Python)
• [x] All updated documentation (see "To Do") has been proofread
• [ ] I have personally used the SFTP protocol to perform a push and a pull on Windows, grabbing from a POSIX machine

PR Type

• [ ] This PR introduces a breaking change (will require a bump in the minor version)
• [ ] The changes in this PR are high urgency and necessitate a hotfix or patch release (will require rebasing off of release)
• [ ] This is a release (staging) PR (maintainer use only)

Checklist:

• [x] I have read the contributor's guide
• [ ] I have run mkdocs serve locally and ensured that all API docs and changes I have made to the static pages are rendering correctly, with all links working
• [ ] All tech debt concerns have been resolved, documented as issues, or otherwise accepted
• [x] I agree to license my contribution to this project under the GNU Public License v3

[OpenBagTwo]

I can't believe I forgot I'd already found (and then just failed to apply) the solution to the whac-a-bug...

[OpenBagTwo]

Well this is interesting... it looks like -p enderchest.test.plugin is completely messing up the coverage report...

For now I appear to have things working by generating the report manually:

coverage run --source enderchest -m pytest --use-local-ssh
coverage report -m

but I need to either find an existing bug report (and hopefully a workaround) or file a bug report.

[OpenBagTwo]

Yeah, this appears to be a known pytest-cov limitation: https://github.com/pytest-dev/pytest-cov/issues/587

[OpenBagTwo]

I'm calling it with regard to unit tests. Even though overall coverage is only up two percentage points (85%), coverage within the sync subpackage is looking good, with the stuff that's not tested (and not "no cover" pragma'd) being stuff that should be tested, but isn't currently worth the level of effort.

[OpenBagTwo]

Trying this for real, some observations:

1. Due to SFTP having to individually scan every single sub-directory on the remote, computing a diff is extremely slow (for my 22GB EnderChest :grin: )
2. Because of the way that dry runs work, it has to perform the diff twice for a typical sync operation (it also prompts you twice for the password). That's not wrong, per se (the diff could change in the pause interval, especially if it's asking for a manual confirmation) but it isn't a great user experience.

[OpenBagTwo]

I performed an enderchest open and used the -v flag to track the progress, but cancelled something like halfway through ('cuz recursing through 22G of Minecraft files is sloooowwww).

I also ran an enderchest close -v --dry-run which produced the expected diff, so I'm feeling good enough about this to merge.

For the release candidate, I'll set up a smaller test chest. And, of course, if anyone reading this dailies Windows and wants to try @dev between now and the RC, I'd definitely love feedback!