test(fm): wait for schedule runs instead of counting them in a fixed window

The schedule tests counted check runs inside 0.1-0.3 s windows. On the CI
runner the failing-check test logged a full traceback per run and fitted
only one run in 0.15 s. Each test now waits (up to 5 s) for the runs it
needs.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>