Python Setup Fix, master branch (2018.01.26.)

Made sure that ${PythonInterp_BINARY_PATH} would be set correctly for the environment setup, after the recent changes.

@fwinkl, I did run some tests. The reason I did was that I honestly remembered the full Athena build to configure itself more quickly than it did from master today. I was suspecting that what we win during the CMake file reading, may in some weird way (because of the new cache variables) slow down CMake's internal calculations. But apparently no...

FindPythonInterp_Measurements.pdf

As you can see the "cache free" code was slower all across the board. (Didn't bother measuring the Ninja case with the old code, as that result really doesn't matter all that much.) Unfortunately the new code doesn't win us any ridiculous amount of time, but it does make the code at least a little bit faster, so let's set up the code like this. (I suspect that most of the improvement comes from not executing a python command over and over again.)