The function that retrieves the dependable job is pretty basic and
retrieves the first job found with the matching name, but this job can
be failed but then successfully retried. In that case, we would exit the
depending job even though the dependable job actually succeeded (the
second time). Let's simplify things, be optimistic and continue with the
depending job even if the dependable job fails.

That reverts to the original behavior.

Signed-off-by: Rémy Coutable <remy@rymai.me>

[Review Apps] Use 2 replicas for the `tiller-deploy` Deployment

Closes #55323

See merge request gitlab-org/gitlab-ce!23806

Exit `wait_for_job_to_be_done` as error if the job fails

See merge request gitlab-org/gitlab-ce!23997

`--no-rdoc` and `--no-ri` have been deprecated since RubyGems 2.0:
https://github.com/rubygems/rubygems/blob/v2.0.0/History.txt#L43

They were finally removed in a recent release:
https://github.com/rubygems/rubygems/pull/2354

If a script is waiting for a job to be done and that job fails,
exit with an error status so that the script doesn't continue
with a prerequisite in an invalid state.

Signed-off-by: Rémy Coutable <remy@rymai.me>

Signed-off-by: Rémy Coutable <remy@rymai.me>

Optionally compiling assets when needed for qa jobs proved
unworkable so instead we fall back on compiling them for
all qa jobs.

CE-EE parity for prepare_build script

See merge request gitlab-org/gitlab-ce!23819