feat(spanner): use single change stream reader via partition queue

[thetechnick]

Replace per-partition goroutine spawning with a serialised partition queue so that at most one Spanner change stream read is active at any time, keeping the process well within the 20-reader-per-stream limit.

Previously, handleChildPartitionsRecord spawned a new goroutine (and a new b.client.Single().Query) for every child partition produced by a split or merge. Multiple concurrent readers of the same change stream cause the Spanner emulator to distribute writes non-deterministically across partitions, making events visible to one reader but not another.

Changes in broadcaster.go:

• Add partitionWork struct and buffered partitionQueue channel (cap 128).
• Add runPartitionLoop: the single goroutine that dequeues and calls readChangeStream one at a time.
• readChangeStream loops continuously, re-issuing the query on iterator. Done with no children (common on the emulator when no data is buffered yet). The emulator buffers writes even when no query is active, so events are never lost between re-queries.
• readChangeStream returns only when sawChildren == true (partition ended due to split/merge), so runPartitionLoop can dequeue and read the children that handleChildPartitionsRecord enqueued.
• handleChildPartitionsRecord enqueues child partitions into partitionQueue (without holding the lock) instead of spawning goroutines. enqueuedChildren set replaces spawnedChildren to prevent duplicate enqueues when the same token appears more than once.

Changes in broadcaster_test.go:

• Rename TestBroadcaster_SingleReader -> TestBroadcaster_PartitionManagement and update it to verify the queue-based bookkeeping (enqueuedChildren, pendingChildren) directly.
• In TestBroadcaster_DeleteEvent and TestBroadcaster_UpdateEvent, move Subscribe before Create so the subscriber is always registered before any events are processed, eliminating a race where the CREATE could be broadcast before the subscriber existed.
• Increase eventTimeout from 10s to 30s: the Spanner emulator takes up to ~20 seconds to deliver DELETE and UPDATE events through the change stream, so a 10-second timeout caused reliable test failures.

[openshift-ci]

Skipping CI for Draft Pull Request.
If you want CI signal for your change, please convert it to an actual PR.
You can still manually trigger a test run with /test all

[openshift-ci]

[APPROVALNOTIFIER] This PR is APPROVED

This pull-request has been approved by: thetechnick

The full list of commands accepted by this bot can be found here.

The pull request process is described here

Details

Needs approval from an approver in each of these files:

• OWNERS [thetechnick]

Approvers can indicate their approval by writing /approve in a comment
Approvers can cancel approval by writing /approve cancel in a comment

[openshift-ci]

PR needs rebase.

Details

Instructions for interacting with me using PR comments are available here. If you have questions or suggestions related to my behavior, please file an issue against the kubernetes-sigs/prow repository.