Generate a PDB that protects SpiceDB replicas

[ecordell]

This prevents infrastructure changes from disrupting more than 1 pod in a SpiceDB cluster at a time.

Fixes CTL-874

[tstirrat15]

I don't feel like I'm expert on PDBs, but the code generally looks good to me.

One upshot of this is that rollouts will take longer because each pod is going to wait until it's up and healthchecking before the next one goes, right?

[ecordell]

One upshot of this is that rollouts will take longer because each pod is going to wait until it's up and healthchecking before the next one goes, right?

That was already the case with the Deployment config we used, but this now also prevents other infrastructure changes from disrupting the cluster. We saw this when we rolled a config change to Nodes that rolled them all; the PDB would have prevented the next node from rolling until the minimum availability was back up