Fix the k8s deploy CI job

[dboreham]

It began failing (with no code changes) some time in the past few weeks.

[dboreham]

Problem vanished as quickly as it appeared. Perhaps a runner timing issue?

[dboreham]

Reopening since the failure is intermittent.

[dboreham]

Working theory is that this is a timing issue. Failure is on ingress creation. There is code to wait for the ingress controller to become ready, which it is, but presumably it's not really ready, quite. Test fails sometimes, passes sometimes.

[dboreham]

I added some exception catching to get rid of the stack trace palooza and tried to reproduce a bunch more times.
It's hit and miss. I can make the error occur about 1 time in 10. Working theory is that on a slower runner the ingress service isn't really up and running when it claims it is, leading to the error when we create the ingress object. That said, there has to be some k8s design error here because just creating the object shouldn't throw a fatal error because some cluster-side thing isn't up and running.

[dboreham]

Hopefully addressed in #116