Prevent concurrent refresh operations from racing in EventManager

[Copilot]

Addresses two issues in the EventManager refresh error handling introduced in PR #886: a race condition where concurrent refreshes could overwrite newer data with stale snapshots, and missing test coverage for the error fallback behavior.

Changes

• EventManager.swift: Added maxPublishers: .max(1) to .flatMap operator to serialize refresh operations and cancel in-flight refreshes when new ones trigger
• FakeEventStore.swift: Added error injection flags (shouldThrowOnFetch*) and call counters to enable testing fetch failures
• EventManagerTests.swift: Added RefreshErrorHandlingTests.test_eventsPreservedOnRefreshFailure() verifying calendars/events retain previous values when fetches fail

Race condition scenario (now prevented)

// Before: Multiple refreshes could run concurrently
1. Refresh A starts, captures current = [Event1]
2. Refresh B starts, succeeds, publishes [Event1, Event2]
3. Refresh A fails, publishes current = [Event1]  // ❌ Overwrites newer data

// After: maxPublishers: .max(1) cancels Refresh A when B starts
1. Refresh A starts, captures current = [Event1]  
2. Refresh B starts, cancels A, succeeds, publishes [Event1, Event2]  // ✅

---

✨ Let Copilot coding agent set things up for you — coding agent works faster and does higher quality work when set up for your repo.