Fix spike waveform transpose issue in Python3

[jonnew]

• Implements the same fix given in OpenEphys.py in main directory, which
  I guess is for Python2.7
• Requesting code review before merge

[jonnew]

As a side note, why are their different versions of these libraries being maintained for Python 2.7 and 3.5 when there are standard mechanisms in place to detect the interpreter being used?

https://docs.python.org/3/howto/pyporting.html

[jonnew]

I created a loadSpikesFast function that results in many orders of magnitude speed up in load times for spikes using numpy's structured arrays. Requesting test and code review before merge. Also, I stopped editing the OpenEphys.py file in the main directory because it seems pretty stupid to maintain two files.

[jvoigts]

This looks very clean and fairly bug proof, unless someone else would like to test this on some other example files i'd say we can just merge this?
Oh, and of course spikes have a color.

[jonnew]

OK, cool. One thing to note is that the return type is different than loadSpikes and I have not found a clean way around this. Basically, the actually data fields are nested ones step lower down in the dictionary than for the loadSpikes data:

data = loadSpikesFaster(filename)
print(data['data']['waveform'])

vs.

data = loadSpikes(filename)
print(data['spikes'])

to get equivalent data access. This is why I created the new function instead of just modifying existing. Ways to do this without breaking code would be best, but ideas I've had are not elegant. Of course this is also a result of returning the header with the data. Perhaps the best way around this would be to have two functions, one that returns the header only, for generic file types, and one that returns only the data. Most generic would be to have two functions getData and getHeader. getData should be able to deduce return type based upon header info and delegate to 'private' functions.