FIX: No longer depending on pydotplus (networkx >=2.0 update)

[TheChymera]

@effigies

Fixes: #3244

didn't commit after running make check-before-commit because it changes a lot of files with what appear to be unrelated changes.
Also it fails with:

chymera@silenthost ~/src/nipype $ make check-before-commit
Checking specs and autogenerating spec tests
env PYTHONPATH=".:" python tools/checkspecs.py
CompCor has nonautotest
TSNR has nonautotest
200917-12:08:45,637 nipype.interface WARNING:
	 VTK was not found
200917-12:08:45,638 nipype.interface WARNING:
	 tvtk wasn't found
ErrorMap has nonautotest
Overlap has nonautotest
Overlap has nonautotest
TSNR has nonautotest
200917-12:08:52,951 nipype.interface INFO:
	 We advise you to upgrade DIPY version. This upgrade will open access to more function
200917-12:08:52,953 nipype.interface INFO:
	 We advise you to upgrade DIPY version. This upgrade will open access to more function
200917-12:08:52,955 nipype.interface INFO:
	 We advise you to upgrade DIPY version. This upgrade will open access to more models
FSSurfaceCommand has nonautotest
BBRegister has nonautotest
FILMGLS has nonautotest
Traceback (most recent call last):
  File "tools/checkspecs.py", line 522, in <module>
    ic.check_modules()
  File "tools/checkspecs.py", line 475, in check_modules
    bad_specs = self.test_specs(m)
  File "tools/checkspecs.py", line 230, in test_specs
    __import__(uri)
  File "/home/chymera/src/nipype/nipype/sphinxext/apidoc/__init__.py", line 5, in <module>
    from sphinxcontrib.napoleon import (
ModuleNotFoundError: No module named 'sphinxcontrib.napoleon'
make: *** [Makefile:73: specs] Error 1

I know I ran into this before, maybe this is another unnecessary dep best fixed in another PR.

Acknowledgment

• (Mandatory) I acknowledge that this contribution will be available under the Apache 2 license.

[TheChymera]

@effigies ^ ?

[effigies]

Pushed a commit to master that I think should fix the test failures. If everything's green, feel free to rebase on that.

[effigies]

Test failures fixed on master.

[TheChymera]

@effigies rebased, but the circleci tests seem to still fail. Any ideas why? They look like permission errors, no idea how that could be related to the PR....

[effigies]

Not sure. My guess is there's something weird with running CircleCI on your own fork, rather than as a PR on nipy. But Travis is passing, so let's go ahead with this. If there's a continued problem with Circle, we can address it when it reproduces on nipy.