[FR]: Investigate using Retry Step action in GHA infra

[ncooke3]

Description

Pretty much all GHA jobs use the retry.sh script to run them again in the case of failure. Instead, there exists a public GHA– retry-step, that gives more granular control over how the step is re-run. This might be particularly relevant for networking related flakes that could benefit from a timed pause between successive runs.

Steps

1. Experiment with retry-step to determine whether it's worth it to switch to it.
2. If it's worth it, delete the retry.sh script and all references to it.
3. Add the retry-step step to all applicable GHA jobs.

API Proposal

No response

Firebase Product(s)

N/A

[paulb777]

After adding a few more in #12371, we got the first green nightly in a while. If it keeps looking good, we should migrate the rest.