fix: skip use-stale background job when cache already refreshed (MR-7760)

[jwadhams]

Summary

• Adds cacheIsCurrentlyFresh() to AbstractUseStaleRequest — returns true when the refresh key holds the 'refresh after' marker written by a successful writeResponseToCache() call
• The dispatched background closure checks this before making a network request; if another job has already refreshed the cache, it returns immediately
• Covers the storm scenario: a backlogged queue drains and many duplicate refresh jobs hit upstream simultaneously — now only the first one through actually makes the network call

Why needsRefresh() alone doesn't work here

needsRefresh() returns false for two distinct states:

State	refreshCacheKey value	needsRefresh()
Job queued, hasn't run yet	'Wait between refreshes'	false
Prior job succeeded	'refresh after'	false

Using needsRefresh() in the closure would incorrectly bail out jobs that run quickly (within the 5-minute snooze window). Checking the actual stored value correctly distinguishes the two states.

Test plan

• testDeferredJobSkipsWhenCacheAlreadyRefreshed — new test: two jobs queued across the waitBetweenRefreshes boundary; J1 runs and refreshes; J2 asserts zero additional network calls
• All 74 existing tests pass unchanged
• CI green

Closes MR-7760

🤖 Generated with Claude Code