chore(ruff): stop tracking ruff/site, the committed build output

[PerryLink]

What

ruff/site/ is the zensical build output of ruff/docs/, and it is committed to the
repository: 13 files, 513,202 bytes (git ls-files -- '*/site/*';
sizes are the blob sizes, as git cat-file -s reports them).

$ git ls-files -- "*/site/*"
ruff/site/404.html
ruff/site/assets/images/favicon.png
ruff/site/assets/javascripts/LICENSE
ruff/site/assets/javascripts/bundle.6e5f0216.min.js
ruff/site/assets/javascripts/workers/search.e2d2d235.min.js
ruff/site/assets/stylesheets/classic/main.a2001754.min.css
ruff/site/assets/stylesheets/classic/palette.7dc9a0ad.min.css
ruff/site/assets/stylesheets/modern/main.fba56155.min.css
ruff/site/assets/stylesheets/modern/palette.dfe2e883.min.css
ruff/site/index.html
ruff/site/objects.inv
ruff/site/search.json
ruff/site/sitemap.xml

ruff/site/objects.inv is 0 bytes. The repository's own ignore rules already say this path
should not be tracked:

$ git check-ignore -v --no-index ruff/site/index.html
.gitignore:1:site	ruff/site/index.html

$ git check-ignore -v --no-index uv/site/index.html
uv/.gitignore:5:/site	uv/site/index.html

• Root .gitignore ignores site at line 1, and /site again at line 158.
• uv/.gitignore:5 repeats /site for that module; ruff/.gitignore has no site rule at all
  (/.venv, /**/__pycache__, /sdk, /.env), which is how this slipped through.
• uv/site/ and deptry/site/ are not tracked, which is the intent everywhere else.

It is not the published site

Pages for this repository is build_type: workflow (served at https://daggerverse.docs.typesafe.ai/),
so nothing is published from a checked-in directory. The Dagger module builds each site inside
a container (_zensical_site, .dagger/src/typesafe_daggerverse/main.py:124-137) and
docs_build assembles ./public from those container outputs;
.github/workflows/deploy-docs.yml uploads ./public, not ruff/site/. Neither that module
nor that workflow reads the committed path.

More directly: a code search for the literal ruff/site across this repository returns
total_count: 0 — nothing in the tree references it.

It is already stale

The committed ruff/site/search.json has six items and no entry for Where to go next, which is
a section of the source file it was built from (ruff/docs/index.md:51). So the tracked copy no
longer describes the documentation that sits next to it in the same commit.

That is the real cost: a reader who opens ruff/site/index.html in the repository sees
documentation that does not match ruff/docs/, and it can only mislead.

Change

$ git rm -r --cached ruff/site

and /site appended to ruff/.gitignore, for parity with uv/.gitignore. The root .gitignore
already covers it, so the deletion alone is enough to stop tracking; the extra line keeps the file
from coming back via git add -f habits and mirrors the sibling module.

Files are removed from the index only — this does not delete anything from an existing working
tree, and no build, test or workflow reads the path.

Verification

• Checked against main @ b0e9d0ca81328d3c91a08ab8c7a0e8963c2fa7ae.
• Tracked file list and byte sizes read from the git tree at that commit (13 blobs, 513202 B).
• git check-ignore -v --no-index output above is quoted from the repository's own rules.
• Pages configuration read from the repository's Pages API (build_type: workflow).

If ruff/site/ is deliberately tracked for a reason I have missed, close this and I will drop it —
but the ignore rules and the Pages setup both point the other way.

[danielgafni]

Sorry I was mass-closing PRs!

[PerryLink]

No worries about the mass-close — thanks for catching it and reopening.

I re-checked this against the current main so you don't have to: the branch is one commit behind (f67568ba, the uv virtual-members / uv_build change), and that commit touches 20 files, all under uv/, .dagger/ and github/. None of them is under ruff/, so there is no overlap with the 14 files here and the PR still merges as-is (mergeable: true). Happy to rebase if you would rather have a linear branch.

The case also still holds on today's main: ruff/site/ is still tracked at 13 files / 513,202 bytes, and ruff/.gitignore still has no site rule — while the repository root .gitignore:1 has site and uv/.gitignore:5 has /site. ruff/ is the one directory of the three that was left out, which is why its committed zensical output ended up in the tree.

Nothing else needed from my side; it can land whenever it suits you.