test: fix fake informer removal race

[outsider987]

Summary

Fixes #462.

Synchronize the test fakeInformer's shared slices, signal completion after removal, and fail explicitly on timeout. Callbacks and channel waits run outside the mutex. Production code and the existing event-count assertions are unchanged.

Verification

Local Linux/amd64, Go 1.26.0; baseline c4b075131e36a1c89b39c89baa7615396d009992.

Check	Result
Original test, -race -count=100	FAIL: data race and handler-removal assertion
Fixed test, same command	PASS
Fixed test, -race -count=100 -cpu=1,2,4	PASS (300 executions)
Entire pkg/relatedresource package, -race	PASS
Full repository, go test -mod=readonly ./...	PASS
Mutation: signal completion but retain the handler	FAIL: existing handler-removal assertion
Mutation: skip removal and its completion signal	FAIL: explicit timeout after 1 second

go test -mod=readonly -race ./pkg/relatedresource -run '^Test_addResourceEventHandler$' -count=100
go test -mod=readonly -race ./pkg/relatedresource -run '^Test_addResourceEventHandler$' -count=100 -cpu=1,2,4
go test -mod=readonly -race ./pkg/relatedresource
go test -mod=readonly ./...

Mutations were tested through temporary Go overlays, not included in this change. This verifies the test helper's synchronization, not a production-cluster failure.

Before / after: synchronization

flowchart LR
  subgraph Before
    B1[Signal completion] --> B2[Remove handler]
    B1 --> B3[Test resumes]
    B2 -.->|Unsynchronized shared slices| B3
  end
  subgraph After
    A1[Lock shared state] --> A2[Remove handler]
    A2 --> A3[Signal completion and unlock]
    A3 --> A4[Test checks subsequent events]
  end

Loading

Before the fix, the waiter's slice lookup can also race with removal, even before it receives the completion signal. The fix protects that lookup and captures the channel under the same mutex.

[outsider987]

Could a maintainer approve the pending CI run and review this test-only fix? Local race checks and the full test suite pass. Thanks!

[brandond]

Thanks for the fix... can I ask why you picked this up? Wrangler doesn't see a lot of use outside the Rancher ecosystem and the issue was created by a maintainer mostly just to track a flake, so I'm curious what brought you here.

[outsider987]

Thanks for asking! Someone I know in the broader SUSE ecosystem mentioned that he’s leaving soon and suggested I keep an eye out for potential opportunities. That’s what got me looking through the related repositories. I wanted to get familiar with the ecosystem and find a small, well-scoped issue I could genuinely contribute to ^_^ Also, just to make sure — does the fix itself look okay, or is there anything I missed? I also used AI as a coding assistant while investigating this, but the PR was based on my understanding of the code and the potential race condition I found. if it not expect pls let me know it!!

[outsider987]

if that is bad I will cancel this fixed, sry

[brandond]

no, no problem - just curious.

[outsider987]

let's go~~~

[brandond]

lgm. someone from @rancher/rancher-squad-frameworks needs to approve as well.