Tests fail more frequent after switch to use Github action

[zhangtbj]

After we switch to use GitHub Action as CI process, we see more frequent test failures than before.

For this monday as an example, there are about 5 PRs, but most of the PRs failed at integration or e2e test steps, but some of them are just the document change. We faced several failures:

• Cluster setup fail
• integration or e2e test fail

such as:

      Expected
          <string>: service account foobar-sa not found
      to equal
          <string>:  \"foobar-sa\" not found

      /home/runner/work/build/build/test/catalog.go:311

Here: https://github.com/shipwright-io/build/pull/641/checks?check_run_id=2070257253

We don't know how to improve it now, but would like to open this issue to record the recent failures then think how to solve the failures better.

So we will continue adding the failure message here then think about how to solve it.

[imjasonh]

Some more flaky-seeming recent test failures:

• Message: " \"default\" not found"
• Expected "default" to be "pipeline"
• error: .status.conditions accessor error: <nil> is of the type <nil>, expected []interface{}

[qu1queee]

For

error: .status.conditions accessor error: is of the type , expected []interface{}

the root cause is that this call hits a nil. Basically at the time the wait is executed, the cbs crd does not have yet a conditions list. I´m not sure on an alternative to address it yet.