Test alphalens.tears.* functions: add matplotlib output comparison

[luca-s]

Currently the tests relative to alphalens.tears.* functions check that no exceptions are thrown. It would be nice to capture matplotlib output, compute a checksum for each figure generated and save those checksums as test references so that we can compare the functions output other than no exceptions are thrown

[luca-s]

In order to accomplish this we need to force a specific matplotlib version for Travis CI, otherwise small differences in the plots due to matplotlib version discrepancies would result in test failures.

Does anybody see a problem in forcing a matplotlib version?

[richafrank]

👍 for reproducible results