Add a post-merge PR deployment cleanup script

[bourgeoisor]

If the PR trigger fails to complete, it may leave a residual deployment namespace (with a running deployment) in the development cluster that will never be cleaned up. These can pile up over time. We should look into having a script (or trigger) that runs periodically and cleans up these namespaces.

[bourgeoisor]

Update: I just deleted hundreds of remnant PR namespaces. We should definitely look into this.

This is what I used:

for namespace in (k get ns -o name | grep pr)
    kubectl delete $namespace
end