webAddressProvenance's no-rescan test is a 50 ms wall clock: it flakes on CI and misses a real rescan #2455

backend/__tests__/unit/services/webAddressProvenance.test.js:154 asserts that 1,000 refused GivenAddresses.has() lookups against three hostile pages finish in under 50 ms, with no warmup. On shared runners that measures the runner: it failed at 62.47 ms on #1768 attempt 1 and passed on attempt 2, and it went red on #2446 as well (both diffs were elsewhere).

The budget is also weak in the direction that matters. Measured: inject the regression the test exists to catch — resolve re-deriving from the stored given texts — but only ONCE per instance, and the clock test passes in 8 ms. A single rescan is invisible to it; only a rescan per lookup is slow enough to trip.

Both problems go away by counting the thing the guarantee is about. A rescan is a count, not a duration.

Found by otto's gate on #1768; row TASK-347 (forge).