Migrate from poetry & tox to uv

[ItsDrike]

This changes our primary dependency management tool from poetry to uv.

Alongside, it also drops tox, as uv doesn't have the equivalent of a poetry-tox dependency, and I didn't really feel like bothering with getting it to work, considering there were plans to remove it anyways (#716). P.S. Sorry for taking this over Kevin, but the PR was stale for a long time now, so hope you don't mind, I did mark you as a co-author on one of the commits, as I pretty much just copied your work on the docs adjustments & poe tasks.

I've already mentioned some of the reasons why uv is likely going to be a better choice for us in the associated issue: #925.

Closes: #925, #488
Superseds & closes: #716

---

Some notes:

• UV doesn't have the equivalent of poetry shell, so venv activation should be done in the standard way, as it would be with the venv module alone (source .venv/bin/activate).
• This will slightly change the process of publishing a package. Right now, we're using a workflow that is triggered on a push of a git version tag (a tag starting with v). We can do this both from the GitHub UI or the CLI. After this is merged, it will also be necessary to change and maintain the version value in the pyproject.toml. Even though I did find an equivalent to poetry-dynamic-versioning, which was doing this for us automatically before: uv-dynamic-versioning, looking at the GitHub stats of this project, it's not very popular, and I'd rather not introduce such a dependency. I did however add a check in the CI that ensures the version in pyproject.toml was in fact bumped before attempting a release (The release would fail anyways as PyPI would refuse it, but this way, it's at least cleaner what happened)
• I've also added another pre-commit check, which makes sure that the UV lockfile is up to date with the pyproject.toml file. This is handy when bumping the project version, as it's easy to forget about running uv lock afterwards, to update project version there too.
• The workflows are using the official action to install UV, which is very fast and provides automatic caching of the global UV cache directory. This means dependencies installed later on within the workflow will be cached through this global uv cache.
• What's a little bit unfortunate is how the uv sync command handles groups, as there's no easy way to say "exclude all groups and only install these ones". Well, there is: --only-group <ONLY_GROUP>, however, it only works with a single group and using it only installs the dependencies from this group, and I do mean only, it doesn't even install the project & it's dependenceis. From what I read in the docs & uv sync --help, I didn't see any way to use this flag & install the project dependencies too. This means that we could either make all of the groups optional, and move the burden up to the contributors installing the project to include the --all-groups or --group <GROUP> for each needed group, or do what I did here, where all the groups are included by default and the CI workflows then exclude those that aren't needed. I will admit that this feels a bit awkward in comparison to poetry there and if we add more dependency groups, it would require updating the commands to exclude those (if desired). I'm fine with making the groups optional instead, if you think that'd be better.
• I have added myself, kevin & perchunpak alongside dinnerbone in the authors section in project metadata. Additionally, I've also included a maintainers section. (Let me know if you want different names / emails used.)
• I've added keywords metadata section for the project, with: minecraft, protocol keywords. (This will show up on PyPI)
• This isn't necessarily something this PR should address, but while I was at it, I've also cleaned up the pre-commit hooks a bit and I've also merged the pyright config into pyproject.toml directly + did some cleanup on it.
• There is a significant speedup in CI runtime compared to poetry: lint takes 20s, rather than 2 minutes. Windows tests finish in about 40s, not 4 minutes. This means we're getting a 6x improvement and that's without cache.
• I've locked docsutils dependency to <0.21, as later versions break the m2r2 parser. See: docutils 0.21 has removed nodes.reprunicode() and therefore building documentation using m2r2 fails CrossNox/m2r2#68. I've just glanced over this, but it seems like m2r2 might be unmaintained, someone in that thread did mention an alternative: myst-parser. It might be worth it to look into this a bit more.
• The publish workflow is currently untested. It should work, but I can't promise it (but in case it wouldn't, we can always fix it and re-attempt the publish, so it's probably not a huge deal)

[PerchunPak]

Great PR, thanks! Just a few nits (mostly)