New design or logic for `test_subprocess.py`

[ccarouge]

A lot of the tests under TestRunCmd in test_subprocess.py are greedy: they test more than they say they are testing.

• test_stdout_is_suppressed_in_non_verbose_mode is supposed to test stdout but it carries assert clauses for stdout and stderr.
• test_stderr_is_suppressed_in_non_verbose_mode is also testing on stdout and stderr.
• test_command_and_stdout_is_printed_in_verbose_mode and test_command_and_stderr_is_redirected_to_stdout_in_verbose_mode are identical tests with different inputs and outputs. Should we parametrize?
• test_output_is_captured_with_capture_output_enabled and test_stderr_captured_to_stdout are the same with different input and outputs. But should the first test only test on stdout and not stderr?
• test_command_is_printed_and_stdout_is_captured_in_verbose_mode: why test on stderr?
• test_stdout_is_redirected_to_file: why test anything else but the file content?
• test_command_is_printed_and_stdout_is_redirected_to_file_in_verbose_mode: why test on stderr?

We could also redesign the tests to be for a series of cases using parameterisation (I'm not sure all are necessary). Tests would be:

• test stdout only
• test stderr only
• test captured stdout (capture_output=True) only
• test captured stderr only
• test file printing only

And run all of these on this series of inputs:

• command only
• command + verbose
• command + capture_output
• command + capture_output + verbose
• command + output_file
• command + output_file + verbose
• same as all the above but with redirection of the command to stderr (echo foo 1>&2)

I'm not sure if all these inputs are necessary or if we need more (do we need to test for capture_output + file printing?).

Also, rename captured in the tests as that is confusing with the capture_output option.