Add an in-memory Bus implementation for tests and cover internal/bus/natsbus publish/close

[Jagadeeshftw]

📌 Description

The bus.Bus interface in internal/bus/bus.go has only one implementation (internal/bus/natsbus/natsbus.go) and no tests. Handlers that publish events (e.g. internal/handlers/github_webhooks.go) can only be exercised against a real NATS server, making them hard to test. There is no in-memory/fake Bus.

💡 Why it matters: A fake bus unblocks fast, hermetic tests for every publisher and documents the contract the real NATS bus must satisfy.

🧩 Requirements and context

• Add an in-memory Bus implementation (e.g. internal/bus/membus) recording published subjects/payloads and supporting Close.
• Cover natsbus.Publish behavior: nil-connection guard and context-cancellation handling, using a connection stub or nats-server test helper.
• Document when to use the fake vs the real bus in PATCHWORK_BACKEND_ARCHITECTURE.md.
• Keep the Bus interface unchanged.
• Wire the fake into at least one handler test to prove the seam.

Non-functional requirements

• Must be secure, tested, and documented.
• Should be efficient and easy to review.

🛠️ Suggested execution

1. Fork the repo and create a branch

git checkout -b test/in-memory-bus

2. Implement changes

• Write/modify the relevant source: new internal/bus/membus/membus.go
• Write comprehensive tests: internal/bus/membus/membus_test.go, internal/bus/natsbus/natsbus_test.go
• Add documentation: architecture doc note on testing the bus
• Include GoDoc comments on the fake
• Validate security assumptions: fake must not be selectable in production builds

3. Test and commit

• Run tests:

go test ./internal/bus/...

• Cover edge cases: publish after close, cancelled context, nil conn
• Include test output and security notes in the PR description.

Example commit message

test(bus): add in-memory bus and cover natsbus publish/close

✅ Acceptance criteria

• In-memory Bus implementation added and tested
• natsbus nil-conn and context-cancel paths covered
• At least one handler test uses the fake bus
• Bus interface unchanged

🔒 Security notes

Ensure the in-memory bus is test-only and cannot be accidentally wired into cmd/api/cmd/worker.

📋 Guidelines

• Minimum 95% test coverage
• Clear documentation
• Timeframe: 96 hours

[xx7412421-cloud]

Hi, I would like to work on this issue if it is still available.

Planned approach:

• Review the existing bus.Bus interface and current natsbus implementation first so the fake preserves the existing contract.
• Add a test-oriented in-memory bus that records published subjects and payloads and handles Close cleanly without changing the production interface.
• Add focused tests for publish-after-close, cancelled context, and nil-connection behavior where the current code supports it.
• Wire the fake into one handler test to prove publisher code can be tested without a real NATS server.
• Document fake-vs-real bus usage in the architecture docs and run go test ./internal/bus/... plus the smallest relevant handler test.

I will wait for assignment before opening a PR so the work follows the GrantFox workflow.