Fixes PSF coeff evaluation

[sbailey]

This PR

• reworks the PSF I/O to fix a bug in the assumptions of the wavelength ranges for XTRACE, XTRACE, and PSF coefficients (they aren't necessarily the same...)
• adds a basic set of PSF coefficient tests
• if specter is installed, the PSF coefficients will be compared to the specter answers; if specter is not installed it will gracefully skip that test

Although the evaluated coefficients now agree, the evaluated spots do not yet agree. The next step will be to add tests that spots agree (they don't) and debug that, then proceed with testing+debugging projection_matrix and ex2d_patch.

The tests are not yet integrated into Travis or CircleCI, but they can be run with

python setup.py test

[lastephey]

Looks good to me @sbailey

+1 for cleaning up the psf bookkeeping +1 for unit tests

Both tests passed-- it seems like we'll need to write some more tests that fail to help us debug. I assume this will be in a future PR. :)

I'll leave it to @dmargala to comment and merge if he chooses