Nested submodules bug and optimization

@svensson This MR:

• fix bug in nested submodules
• optimize esrftaskgraph by disabling persistent task output

However the esrftaskgraph is still a factor 2 slower than bare pypushflow (like you reported before):

pytest esrf2pypushflow/tests/test_ppf_workflow10.py --durations=0

184.40s call     esrf2pypushflow/tests/test_ppf_workflow10.py::test_workflow10[5000]
38.90s call     esrf2pypushflow/tests/test_ppf_workflow10.py::test_workflow10[1000]
19.31s call     esrf2pypushflow/tests/test_ppf_workflow10.py::test_workflow10[500]
8.56s call     esrf2pypushflow/tests/test_ppf_workflow10.py::test_workflow10[200]
4.65s call     esrf2pypushflow/tests/test_ppf_workflow10.py::test_workflow10[100]
1.31s call     esrf2pypushflow/tests/test_ppf_workflow10.py::test_workflow10[10]

pytest pypushflow/test/test_workflow10.py --durations=0

94.28s call     pypushflow/test/test_workflow10.py::TestWorkflow10::test_workflow10_5000
18.61s call     pypushflow/test/test_workflow10.py::TestWorkflow10::test_workflow10_1000
9.52s call     pypushflow/test/test_workflow10.py::TestWorkflow10::test_workflow10_500
4.10s call     pypushflow/test/test_workflow10.py::TestWorkflow10::test_workflow10_200
2.64s call     pypushflow/test/test_workflow10.py::TestWorkflow10::test_workflow10_100
1.08s call     pypushflow/test/test_workflow10.py::TestWorkflow10::test_workflow10_10