Add javascript tests for the template control logic

[tepickering]

Follow-up to the continuous mode fix in aed2c63. That was a browser-side bug in code with no test coverage at all, so nothing caught it before it reached the telescope in the middle of an MMIRS night.

What this adds

A small jsdom suite covering the control logic in wfs.html and cwfs.html, run by node's built-in test runner.

• wfssrv/tests/js/harness.js — mounts a template's control script in jsdom with a stubbed fetch and a virtual clock
• wfssrv/tests/js/controls.test.js — 12 tests
• package.json / package-lock.json — one devDependency, jsdom
• .github/workflows/js-tests.yml — runs npm ci && npm test on node 22 and 24
• CLAUDE.md — documents how to run the JS suite, plus a Deployment section recording that the summit installs from GitHub master (pip install git+... --upgrade into its mmtwfs conda env), so a fix is not deployed until it is pushed to master

Design notes

The javascript is read out of the templates, not copied. The harness pulls the inline <script> block that defines controlButtons out of the .html file and evals it. A copy in the test tree would drift and quietly stop catching regressions.

The DOM fixtures are hand-maintained, and guarded. A test scrapes every getElementById('...') id out of each template's script block and asserts the fixture provides it, so a renamed or newly added control fails loudly instead of silently going untested.

Timers run on a virtual clock. The 1s poll loop and the 5s running-flag reset fire in the correct relative order without the tests waiting on them. The whole suite takes about half a second.

Coverage

• continuous mode analyzes a new frame while the controls are disabled — this is the regression
• continuous mode applies focus/coma/M1 corrections when connected, and none when not
• continuous mode keeps polling but does not re-analyze the same frame
• continuous mode picks up each new frame as it arrives
• the MMIRS layout (no #mode selector) omits mode= from the analyze URL; a layout with one passes it through
• starting continuous backs the gains off to 0.2/0.5, and Turbo leaves them alone
• stopping continuous restores the gains, re-enables the controls, and halts the loop
• the cwfs Latest button loads the two newest frames and analyzes them

Validation

Run against the pre-fix templates (7bb3450), 7 of the 12 fail, including every continuous mode test. The 5 that pass are the ones that don't depend on the bug — gains, stop, cwfs Latest, and the two fixture guards. All 12 pass on current master, on node 22 and node 26.

Scope

Node is only needed to run these tests. The served page still has no build step, tox.ini and the pixi config are untouched, and nothing here ships in the wheel. Browser-level behavior — bootstrap, mpl.js, websockets — remains uncovered; that would need Playwright and a much heavier CI job.

🤖 Generated with Claude Code

[codecov-commenter]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 32.71%. Comparing base (aed2c63) to head (cf5c340).

Additional details and impacted files

@@           Coverage Diff           @@
##           master      #24   +/-   ##
=======================================
  Coverage   32.71%   32.71%           
=======================================
  Files           2        2           
  Lines         590      590           
=======================================
  Hits          193      193           
  Misses        397      397           

☔ View full report in Codecov by Harness.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.
• 📦 JS Bundle Analysis: Save yourself from yourself by tracking and limiting bundle sizes in JS merges.