Handful of isort changes and fixes

[matthewturk]

This reformats a few of the files that were still failing, and it
changes the setup.cfg section to be isort rather than tool:isort, which
seems to be the right thing for isort 4.3.

[neutrinoceros]

left two comments to further improve stability of isort results.
Thanks for cleaning up this mess !

[neutrinoceros]

So @cphyc showed that there are yet again inconsistent results with isort on travis (see #2759)
This is because I ran isort 4.3.21 locally, while travis is running 4.3.0 (I didn't pin the last digit and didn't anticipate pip would read it as 4.3.0). Meanwhile, I've also tried the latest release 5.1.2 (only a few hours old !) and it fixes even more bugs (most notably, it now supports Cython files !) so I think we should simply pin this one.

note that, starting from isort 5.0, the -rc (recursive) option is implicit so it can be removed.
@matthewturk , do you agree to make those infrastructure changes in this PR ? We can open a follow up PR with the new isort pass afterwards.
here's what's needed:

• update tests/lint_requirements.txt to pin the latest version
• update .travis.yml and .github/PULL_REQUEST_TEMPLATE.md (remove the -rc flag)

edit: isort 5.1.3 just came out, looks like it's actively under development for bugfixes these days, so we probably don't want to pin an exact version for now.

[matthewturk]

Went ahead and updated as per your suggestion!

[neutrinoceros]

you should also remove the inline comment in that file (might need to rebase this branch :/)

[matthewturk]

So, unfortunately, I'm struggling to get this to all work. Can we take this as-is and remove the comment in another PR?

[matthewturk]

(As in, specifically, getting the git commit hook to work with the version of isort that also works here etc etc)

[neutrinoceros]

So, unfortunately, I'm struggling to get this to all work. Can we take this as-is and remove the comment in another PR?

sure, works for me

[cphyc]

LGTM!

[neutrinoceros]

note that linting is expected to fail here, and is fixed by the companion PR #2764

[neutrinoceros]

In the end it turns out this branch isn't necessary after tests passed in #2764, so as we agreed on slack I'm closing this :)

[neutrinoceros]

(turns out there's no "close" button on the mobile app yet !)