Add CISUPPORT_PARALLEL_PYTEST variable

      # 10 GiB should be enough for just about anyone :)

      ulimit -m $(python -c 'print(1024*1024*10)')

     # Not (Gitlab and GPU CI)?

     if [[ $CISUPPORT_PARALLEL_PYTEST == "" || $CISUPPORT_PARALLEL_PYTEST == "xdist" ]]; then

       # Default: parallel if Not (Gitlab and GPU CI)?

       PYTEST_PARALLEL_FLAGS=""

       if [[ $CI_RUNNER_DESCRIPTION != *-gpu ]]; then

         PYTEST_PARALLEL_FLAGS="-n 4"

       fi

     elif [[ $CISUPPORT_PARALLEL_PYTEST == "no" ]]; then

         PYTEST_PARALLEL_FLAGS=""

     else

       echo "unrecognized scheme in CISUPPORT_PARALLEL_PYTEST"

     fi

      ${PY_EXE} -m pytest \

        --durations=10 \