whitelist mpl_toolkits tests

[WeatherGod]

Turns out that the tests for mpl_toolkits needed to be explicitly added, as well as a need to explicitly install the baseline images as well.

Note that Travis will fail until the trisurf3d fix is applied to master.

[tacaswell]

@mdboom @efiring Is there a reason these tests are not currently included?

[efiring]

@tacaswell, I don't recall any discussion of this; but I was not directly involved in any of the test system development. My concern would be that if Travis fails every time a change in the mpl core breaks something in 3D, for example, our already slow maintenance and development could grind to a halt. One hopes this would not occur in practice, of course.

[mdboom]

I don't remember why they weren't included. Oversight is the best I can think of. I agree with @efiring in general, but I think if we're saying we reasonably guarantee mpl3d works (which we do, given its limitations), we probably should be testing it. Speeding up the tests -- well -- that's a whole other ball of wax and maybe doesn't need to hold this up.

[WeatherGod]

I force pushed a rebased version of this branch to include all of the latest fixes. I expect it to fail for the two py3k runs. I'll have a PR momentarially for that.

[WeatherGod]

Just did another rebase and force push. Travis is running again.

[WeatherGod]

Test pass! Anybody want to merge?