Add /health handler to node server

[tillrohrmann]

This commit adds a very simple /health handler that returns 200 if the server is running. This allows to check whether the node server is running.

This fixes #2461.

[AhmedSoliman]

The failing test is a little concerning. Is this something that @pcholakov should look at?

[AhmedSoliman]

I feel that I'm missing a little context on why this is needed, and if grpc's health check endpoint can be a reasonable replacement or not.

[tillrohrmann]

The failing test case is captured here #2428.

[pcholakov]

I'm on it; it's very odd - the only way I can explain what we're seeing is that occasionally the object_store file:// protocol provider silently fails to either write or read back the data we've recently stored. I haven't been able to reproduce this yet.

[tillrohrmann]

The motivation for this PR was that we were using in our tests (verification, e2e, local cluster) the ingress and/or admin http /health endpoint to await that the binary is up and running. With the provision step, this is not always possible anymore because we only start those components after the provisioning. That's why I used as a band aid the /metrics endpoint to await the start of the node server in those tests.

I can look into the grpc health endpoint if you think that this would be more suitable.