Py23 compatibility

[evanbiederstedt]

Compatibility with both 2.7 and 3.x.
Encoding/decoding strings in order to bypass this issue: h5py/h5py#289

Revisions to travis config, README, style changes

[evanbiederstedt]

RE: README

As I mention, I rather prefer the motivation/details for the first few sentences in the README. I re-implemented some of the previous old README sentences, as it gives an explanation of Classic HotNet and HotNet2.

Is it clean to users given the README what num_cores=-1 means? We should probably make this more clear.

RE: C/Fortran

In principle, users should compile these extensions when installing the python package, similar to WExT. I haven't dealt with this. I don't think it is very important for the use case. I've kept setup_c.py and setup_fortran.py separate.

RE: py23 compatibility changes

--- Any .iteritems() became .items()

--- We were previously using implicit relative imports. Python does not like this----these need to be explicit relative imports

--- https://github.com/raphael-group/hotnet2/blob/master/scripts/permuteNetwork.py#L70

cores = mp.cpu_count() if args.cores == -1 else min(args.cores, mp.cpu_count)

I actually think this was a bug. Even in Python2.7, this should be mp.cpu_count(), not mp.cpu_count

I should note that variations of this line of code are used at every instance of multiprocessing, but we don't explicitly check for min(args.cores, mp.cpu_count) each time. I'm torn whether this check is necessary. I've therefore left these lines alone.

--- Tuple parameters to functions is not allowed in Python3.x. I unpack the tuples within the function, which is both 2.7 and 3.x compatible.

e.g. https://github.com/evanbiederstedt/hotnet2/blob/master/hotnet2/permutations.py#L13-L14

--- The largest change is the work around for load_hdf5() and save_hdf5()::

https://github.com/evanbiederstedt/hotnet2/blob/master/hotnet2/hnio.py#L391-L455

I save the HDF5 as type S, and decode to be type np.unicode_ upon loading the HDF5.