Benchmark tests

[Mossaka]

While the integration test pipeline creates a k3d cluster for testing end-to-end workloads, we don't put any stress on the cluster to test the performance of the spin shim, and thus have no immediate response to performance degration of PR changes.

I propose to add a separate gh workflow to run some benchmark tests on the cluster:

1. Load test: using tools like wrk or hey to load test the HTTP server and report result as job annotations
2. Baseline comparison: run spin apps on runc shim vs. spin-shim
3. Gradual test: gradually scale out the pods and monitor the cpu, memory, disk and network io usages.

Related: containerd/runwasi#97

Feel free to add ideas and thoughts on this topic! Any suggestion is welcome 🙏