chore(spanner): integrate CacheSubscriber into DatabaseClient lifecycle

[olavloite]

Initializes and starts the proactive background CacheSubscriber when location-aware routing is enabled on DatabaseClient. Ties background subscriber task termination to the RAII Drop lifecycle of LocationRoutingState, ensuring gRPC stream cleanup when all DatabaseClient clones are dropped.

[gemini-code-assist]

Code Review

This pull request integrates the background CacheSubscriber into the DatabaseClient lifecycle when location-aware routing is enabled, refactoring LocationRoutingState to manage the subscriber and delegating database ID retrieval to CacheUpdater. A critical memory visibility issue was identified where replacing the atomic load of database_id with CacheUpdater::database_id() downgrades the memory ordering from Acquire to Relaxed. This breaks release-acquire synchronization and could cause threads to read stale cache entries. It is recommended to update CacheUpdater::database_id() to use Ordering::Acquire to restore proper synchronization.

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 96.59%. Comparing base (836e762) to head (08d4c0e).

Additional details and impacted files

@@           Coverage Diff           @@
##             main    #6573   +/-   ##
=======================================
  Coverage   96.59%   96.59%           
=======================================
  Files         304      304           
  Lines       91987    91909   -78     
=======================================
- Hits        88855    88780   -75     
+ Misses       3132     3129    -3     

☔ 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.

[olavloite]

/gemini review

[gemini-code-assist]

Code Review

This pull request integrates the background CacheSubscriber into the DatabaseClient lifecycle when location-aware routing is enabled. It removes the separate database_id atomic field in LocationRoutingState in favor of retrieving it directly from CacheUpdater, which now uses Acquire ordering for loading the database ID. The PR also refactors mock tests to use a helper function create_test_mock to reduce boilerplate, and adds new lifecycle tests for the cache subscriber. There are no review comments, and the changes look solid, so I have no feedback to provide.