python3 updates

[sbailey]

This PR makes a bunch of little changes for python3 compatibility. The code works for both python 2.7 and 3.5, so I'll try to support both for now, though python 2.7 support could be dropped without warning if some incompatibility arises. I think I've configured Travis to test both as well.

A few notes:

• in a few test scripts, I chose to do an absolute import specter.blah import instead of a relative import ..blah so that the test script could be run by itself for faster focused testing of just that single piece.
• test_binscripts.py is the only place that needed a special case for python 2.7 vs. 3.5 for the logic of where the script is for when it is run; this was needed for finding the bin scripts to test them without installing them first. This is handled as a special case, and is the only 2.7 vs. 3.5 special case needed here.
• most bugs that were not caught by 2to3 were related to integer vs. float division // vs. /
• the other tricky one was isinstance(np.int64(1), int) is True in python 2.7 and False in python 3.5; this was fixed using isinstance(np.int64(1), numbers.Integral) instead which is True for both python 2.7 and 3.5.

[weaverba137]

This all looks good to me, but for maximum assurance that 2.7 & 3.5 behave the same, you should also add unicode_literals to the from __future__ statements. This is particularly important if you do any reading or writing of string data (i.e. non-numeric text) from files.

[sbailey]

I added unicode_literals throughout except for bin/specter, since numpy dtypes doesn't work with unicode under python 2.7 (but it does under 3.5). e.g. this works for python 2.7 without unicode_literals and python 3.5 either way:

x = np.zeros(10, dtype=[('BLAT', 'f8')])

but adding "from future import unicode_literals" breaks it on python 2.7 (though it continues to work on python 3.5).

If really needed, I could use str('BLAT') to get away from unicode, but that is basically just working around a bug in numpy that doesn't happen unless unicode_literals is invoked anyway.