fix: stop PeriodicScheduler from re-arming itself after stop()

[davidberenstein1957]

What changed

PeriodicScheduler.start() now checks _stopped on the reschedule path too (inside the lock), _run() re-checks the flag before invoking the payload, and stop() reads the flag under the lock.

Why

_run() called start(from_run=True), which deliberately bypassed the _stopped check, so a fired timer always armed a successor. A stop() landing between the timer firing and _run() reaching start() cancelled an already-fired (uncancellable) timer and was then silently undone — the scheduler kept firing forever. The payload also ran to completion after stop() returned, so _measure_power_and_energy could mutate tracker state and call live_out after EmissionsTracker.stop() had written the final row and exited the handlers.

The elif not self._stopped: return branch preserves today's "a second start() on a running scheduler is a no-op" semantics. The payload deliberately stays outside the lock, so stop() never blocks for a whole measurement (and must not be called from the payload itself).

Verification

New tests/test_scheduler.py. test_stop_during_timer_firing_does_not_rearm fails on master (scheduler keeps firing, _stopped flips back to False) and passes here; the other three cover stop-mid-payload, double start() arming one timer, and stop() before start() / twice. tests/test_emissions_tracker.py passes unchanged.

Closes #1308

🤖 Generated with Claude Code

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 91.44%. Comparing base (eb2cfef) to head (f562682).

Additional details and impacted files

@@            Coverage Diff             @@
##           master    #1324      +/-   ##
==========================================
+ Coverage   91.43%   91.44%   +0.01%     
==========================================
  Files          49       49              
  Lines        5057     5063       +6     
==========================================
+ Hits         4624     4630       +6     
  Misses        433      433              

☔ 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.

[davidberenstein1957]

Closing in favour of #1339, which fixes this same re-arm bug plus the re-entrancy problem in the same file. Two open PRs rewriting scheduler.py is review burden for no extra benefit, so it's better to have one.