build(docs): gitignore generated docs/book/theme/lang-switcher.js

[singlerider]

Summary

• Base branch: master (all contributions)
• What changed and why: xtask::inject_lang_switcher_locales rewrites docs/book/theme/lang-switcher.js on every cargo mdbook build. Because the file was tracked, any local docs build dirtied the working tree. git add -A then swept it into unrelated commits — several recent PRs carried accidental LOCALES reorders for exactly this reason. This PR removes the file from the index and adds it to .gitignore alongside the other generated docs artifacts.
• Scope boundary: Does not change the xtask injector, locales.toml, or any mdBook build behavior. The file is still emitted at build time; it is simply no longer tracked.
• Blast radius: None. lang-switcher.js is a build artifact. Consumers of the docs build are unaffected.
• Linked issue(s): N/A

Validation Evidence (required)

• Commands run and tail output:

git ls-files docs/book/theme/lang-switcher.js
# (empty — file is no longer tracked)

git check-ignore -v docs/book/theme/lang-switcher.js
# .gitignore:9:docs/book/theme/lang-switcher.js

cargo mdbook build
# lang-switcher.js emitted by xtask injector, build completes normally

• Beyond CI — what did you manually verify? Confirmed the file is untracked post-merge and that cargo mdbook build still regenerates it correctly from locales.toml.
• If any command was intentionally skipped, why: cargo fmt, cargo clippy, cargo test — docs-only change (.gitignore entry + git rm --cached). Replaced with docs quality gate per template guidance.

Security & Privacy Impact (required)

• New permissions, capabilities, or file system access scope? No
• New external network calls? No
• Secrets / tokens / credentials handling changed? No
• PII, real identities, or personal data in diff, tests, fixtures, or docs? No

Compatibility (required)

• Backward compatible? Yes
• Config / env / CLI surface changed? No

Rollback (required for risk: medium and risk: high)

git revert <merge-sha> re-tracks the file. The xtask injector will overwrite it on the next cargo mdbook build regardless.

[Audacity88]

Reviewed current PR head ebf755b. I checked the PR body, labels, CI, comments/reviews, the diff, docs/book/book.toml, xtask/src/cmd/mdbook/build.rs, and a clean scratch checkout of this PR. CI is green and I did not find any prior reviews or review comments.

🟢 Good direction — generated docs churn should not be source of truth

The goal makes sense. lang-switcher.js is partly generated from locales.toml, so removing accidental locale-order churn from unrelated PRs is a useful cleanup.

🔴 Blocking — clean docs builds fail without the tracked JS file

In a clean checkout of this PR head, docs/book/theme/lang-switcher.js is absent from the tree, but docs/book/book.toml still lists it in additional-js. The mdBook build then fails before producing docs:

ERROR Rendering failed ... failed to open 'theme/lang-switcher.js' for hashing ... No such file or directory

The xtask injector does not create the file when it is missing; inject_lang_switcher_locales() returns early if theme/lang-switcher.js does not exist. That means the validation can pass only if an ignored local copy is already present from an earlier build, while a clean clone / docs deploy path still breaks.

Could you keep a tracked template/source file, generate the full lang-switcher.js before mdBook runs, or split this into a tracked source template plus ignored generated output so clean checkouts still have the asset mdBook requires?