[RFC] Use Conda for CI and fix deprecations

[sjkelly]

This builds upon #63. The docker container used for CI was using Julia 0.6 so I adjusted it to now use Conda as well. In addition there are several deprecation fixes for PyCall.

There are some outstanding issues:

• UnitTetrahedronMesh and others seem unexported into the Python FEniCS library, but still exist in the CPP code.
• MPI is not supported in the Conda build and was disabled temporarily
• If installing via Conda should we fix the fenics version to ensure API compatibility?

Should fix #61, #52, and help #28 and #12.

[ChrisRackauckas]

How does this do on Mac and Windows?

[sjkelly]

FEniCS AFAIK is not supported on windows natively. It supposedly runs in WSL. I can test this later to confirm. I guess there should be a warning during the build step to notify users.

[sjkelly]

OSX seems fine with Travis.

[ChrisRackauckas]

Alright, seems fine with me if @ysimillides likes it. The main reason we switched to the Docker was because the original installation method was difficult to deal with. But if this is simpler, then yeah, awesome. Just put a warning for Windows.

[ysimillides]

Hey everything looks really good to me! Just a minor issue, we might want to include the newer julia versions in the testing script.

[ChrisRackauckas]

Yeah, that would probably be a good idea. I don't think it's blocking to this PR though.

[sjkelly]

Thanks! I managed to slip in my sys -> Sys fix with not a second to spare :)