Upgrades

[rebeccacremona]

This PR upgrades Django, and also upgrades any other packages permitted by our constraints specified Pipfile. See Pipfile.lock for a complete list. Of note:

• This commit has landed in django-ratelimit. So, that plus the new setting I added to settings_common.py means we should no longer run into the situation where perma-dev.org goes out of commission due to overstepping our redis instance's limits.

• Upgrades to pytest and pytest-django went too far for pytest-django-ordering; I have a PR out to upgrade the package, but in the meantime have pinned us to use my fork.

• The most recent version of fakeredis is a release candidate, which is too bleeding edge for our code, so I've pinned us to only accept minor and point releases for now.

Everything else seems to have upgraded without problems.

NOTE FOR DEPLOYMENT: I ran into some trouble with pipenv hanging when generating this new lockfile, possibly due to user error re: the package we are installing from my fork. There's no reason there should be trouble, but if anything funny happens, I recommend deleting any pip cache that's on the minion and trying again, and/or throwing out the virtualenv and starting fresh.

[rebeccacremona]

Yeah, this failed on Travis in the way I initially saw locally....... I think it's a virtualenv problem, investigating.....

[rebeccacremona]

The failure was due to the wrong version of Django being installed.

The wrong version of Django was being installed because of the combination of pip 9.0.1, pipenv installing the dependencies in synchronous mode, and installing my forked version of pytest-django-ordering from github using setup.py, which mentions that it requires django but correctly does not specify what version.

Upgrading pip>=9.0.2 was sufficient to solve the problem; running with pip==9.0.1 and pipenv --sequential also solves the problem.

[coveralls]

Coverage remained the same at 83.31% when pulling 58f1fe9 on rebeccacremona:upgrade-django into 56c1121 on harvard-lil:develop.