Lava callback latency problem

Right now callback is fastapi, which is async. But callback endpoint is slow, as it need to upload artifacts to kernelci storage.
Even i did threads, it seems still slow. We have grafana monitoring that does plain HTTP request to default / endpoint, and it is
reaching 2-4 seconds latency when we have lot of LAVA submissions.
We need to investigate this, and probably rewrite way how callback works fundamentally.
**It might be also totally different issue (for example docker-proxy or nginx), because k8s deployment (production) handling this well, but staging (based on docker containers) is slow.**

Start by tracing the FastAPI callback path, including artifact uploads to KernelCI storage, and compare the staging Docker deployment with the production Kubernetes deployment. Use the Grafana request to the default / endpoint as the latency signal, then investigate whether the bottleneck is the callback, docker-proxy, or nginx. Done means the latency cause is identified and a concrete callback or deployment fix is validated.

Written by the indexing model from the issue text.