review CAS memory handling

[migmartri]

The CAS of the running instance of Chainloop has been reaching the limit of 512mi and being killed by the scheduler making some release jobs fail.

We have increased the memory limit to 1GB but we should look into the memory handling of uploads to make sure there is proper handling and no leaks.

[migmartri]

After deploying this release we've seen a memory usage climb

It's probably related to the fact that now the controlplane calls the infoz endpoint of the CAS for its readiness probe

{"level":"info","ts":1681507135.9057927,"kind":"server","component":"http","operation":"/cas.v1.StatusService/Infoz","args":"","code":0,"reason":"","stack":"","latency":0.000021218}

That endpoint seems quite simple, not sure in which part of the stack we are getting that potential memory leak