Include PYTHONPATH in get_env_with_venv_bin for use in subprocesses

[sphuber]

Fixes #4051

The test tests.cmdline.commands.test_process:test_pause_play_kill
was failing in local setups with a time out. The problem was that the
daemon, launched as a subprocess in the setUp, would fail to load the
submitted process, because the class could not be imported. This is
because it is defined in the tests.utils module, which would not be
part of the Python path of the subprocess.

The tests module is automatically added to the path by pytest,
however, it was not being forwarded to the subprocess. Passing the
current environment through the env argument solves this. The only
additional change is to include the PYTHONPATH variable in the env
copy created by aiida.cmdline.utils.common.get_env_with_venv_bin.
These two changes guarantee that the test process classes are importable
by the daemon that is running in the subprocess.

[sphuber]

Only thing that I am not sure about is that this change will also be used in non-test code, since we use the get_env_with_venv_bin function also in normal use. Should we put a switch in to only include the python path when running tests? Normally, these switches tend to be very dirty. Trying to think if it is bad to set the PYTHONPATH for the subprocess from the main. In principle that is the goal of this get_env_with_venv_bin function. We want to make sure that daemon runners have the same environment. So I don't think this should be problematic

[codecov]

Codecov Report

Merging #4052 into develop will not change coverage.
The diff coverage is n/a.

@@           Coverage Diff            @@
##           develop    #4052   +/-   ##
========================================
  Coverage    78.45%   78.45%           
========================================
  Files          461      461           
  Lines        34077    34077           
========================================
  Hits         26731    26731           
  Misses        7346     7346           

Flag	Coverage Δ
#django	70.49% <ø> (ø)
#sqlalchemy	71.35% <ø> (ø)

---

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update f7ac506...8320f99. Read the comment docs.

[sphuber]

After discussion with @giovannipizzi we opt to only apply the change to the test itself, until we properly understand the full effect of always including the PYTHONPATH when copying the env.