chore: deployment startup ordering

[4ndrelim]

1. xtramcp backend needs to be started and ready before paperdebugger backend attempts connection. Currently, this is manually ensured. We can enforce an ordering with a startup gate by including in the initContainer of paperdebugger.
2. [WIP] Add healthcheck node and re-try logic for for more robust handling of connection re-try. This will also facilitate re-establishing connection during pod failures.

Additional:

• Propose to remove defaults for some secrets and allow backend processing logic to handle instead.