Respect pytest -s flag, streaming stdout/stderr as the test runs

[jchorl]

Checklist

• Pull request details were added to HISTORY.rst

I took a crack at #199

Doing some research, it seems pytest already has a setting to dictate how to handle stdout/stderr: https://docs.pytest.org/en/latest/how-to/capture-stdout-stderr.html#setting-capturing-methods-or-disabling-capturing

I added support for -s (which sets capture=no). It bypasses capturing stdout/stderr, streaming them to the terminal in real-time. I did not attempt to implement the other options because they're more complex - capturing and streaming output simultaneously might require a multithreaded implementation. I couldn't find a great way to support it.

Note that this is incompatible with stdout: and stderr: tests (as there is no captured stdout/stderr to check). I wasn't sure where the best place to catch this incompatibility was.

At first, I de-registered the stdout and stderr ContentTestCollector altogether when -s is specified. But this just silently drops those tests - no good.

Then I moved to validating at test runtime in runtest. But ContentTestItem specifies a repr_failure message which didn't really describe this failure mode.

So then I moved to validating in the per-test-item constructor. This does work. I'm not sure if it's the best place.

I tried to add some tests to make sure this functions as expected. I also used variants of the following yaml to manually test:

- name: tee test
  command: bash -c "echo foo | tee test.file && sleep 3"
  files:
    - path: test.file
      contains:
        - foo
        - bar
  stdout:
    contains:
      - bar

Let me know if you think there's a better way to implement this.

[rhpvorderman]

Thank you very much for this. I will look into this when I have the time. I am currently very busy.