Point the stale web/game-worker.js references at tak's worker

[dmccoystephenson]

Summary

The open-issue backlog was empty at triage, so a Stage A documentation accuracy sweep was run. In-repo comments were found that still describe the web runtime as it was before the move onto tak was finished in #197:

• web/game-worker.js is referenced, but it was deleted in Move onto tak (5/5): the web and Pyodide front-ends, and the web/ runtime #197. The Worker is now shipped by tak (tak/web/assets/game-worker.js) and is loaded through /tak/boot.js in web/index.html. The old path was still named in four comments: Dockerfile, src/config/config.py, and two in tests/ui/test_pyodideUserInterface.py. Each is repointed at tak's game-worker.js. The makeSyncSaves() function, the exec of the entry point, and the jsonschema loadPackage call were each confirmed to exist in that file.
• The Dockerfile's reason for installing tak was stale. The note said tak is installed "not because the server needs it" and that "the server here is pure stdlib". web/serve.py is now from tak.web.serve import main, and the kit's assets are served from it at /tak/. The bullet is rewritten to say so. Importing tak.web.serve was checked and was found not to load jsonschema, so the claim that jsonschema is only needed in the browser is kept.
• requirements.txt was left saying FishE "is being moved back onto" tak "one module at a time". That move was finished with Move onto tak (5/5): the web and Pyodide front-ends, and the web/ runtime #197 (5/5). The line is changed to say FishE runs on tak, which is how the README already describes it.

No behaviour is changed. Only comments are edited, so no tests were added.

No other drift was found by the sweep. These were checked:

• The README's Features and Contributing sections: menu labels, fishing hours and energy cost, the $10,000 goal and the $1,000 reveal, export markets and hold capacities, the hauling bonus, the housing ladder, investment property names, FISHE_SAVE_DIR, damaged-<date>-<time>, /tak/ assets, and the SDL handling in run.sh/test.sh.
• schemas/player.json, stats.json and timeService.json against their *JsonReaderWriters. Every written field was found in its schema, and every schema field was found to be written.
• PLANNING.md.

Test plan

• Full suite run locally under SDL_VIDEODRIVER=dummy SDL_AUDIODRIVER=dummy: 910 passed
• Searched for web/game-worker, web/client and being moved: no matches remain
• CI green on the PR head

No tracking issue exists. The gap was found during triage, when the open-issue backlog was empty.

This PR description was drafted during a Gardener session (https://github.com/Stephenson-Software/gardener).

🤖 Generated with Claude Code

---

drafted by Claude on behalf of Daniel Stephenson

[dmccoystephenson]

Self-review rubric (the diff was read in full after CI went green on the head commit):

• Scope: PASS. Four files were touched, and every hunk is a comment naming the deleted web/game-worker.js or describing the unfinished tak move. No code, formatting or rename churn is included.
• Tests-new: PASS (not applicable). No public functions were added.
• Tests-fix: PASS (not applicable). No bug is fixed. The two edits in tests/ui/test_pyodideUserInterface.py are comments only, and no assertion was changed.
• Sibling structure: PASS. No new files were created.
• Sibling renames: PASS (not applicable). No identifiers were renamed.
• Docs: PASS. The README already says "FishE now runs on" tak, and requirements.txt now matches it. No schema, PLANNING.md or version.txt update is warranted for a comment-only change.
• Issue resolution: PASS (not applicable). No Closes #N is claimed.
• CI: PASS. The test check passed in 1m10s on the PR head. 910 tests also passed locally under the dummy SDL drivers.
• Schema-sync / Money-format / Deterministic-tests / Headless-pygame / camelCase: PASS (not applicable). No persisted fields, money displays, tests, pygame paths or identifiers were changed.

Findings:

• Dockerfile:12-13: the claim that jsonschema is loaded by tak's game-worker.js and that the server "never imports it" was verified against the tak installed in this environment, which reports version 0.1.1. requirements.txt pins v0.2.0, and that tag was not inspected directly. The previous comment made the same claim, and nothing in Pin tak v0.2.0 (Pyodide 0.29.5 in the browser build) #198's description suggests the worker's package loading changed. Even so, the claim is marked UNVERIFIED against the pinned tag. It is low-risk because only a comment is affected.
• tests/ui/test_pyodideUserInterface.py:318: the comment line grows to about 81 characters. That is still within black's 88-character limit, and black does not reflow comments, so it is left as is.

Summary: this is a comment-only drift fix following the tak move (#197). CI is green and no protected paths are touched (schemas/, saveFileManager.py, *JsonReaderWriter.py, .github/workflows/, and no single-file deletion over 50 lines).

This review was drafted during a Gardener session (https://github.com/Stephenson-Software/gardener).

---

drafted by Claude on behalf of Daniel Stephenson