Add fitness effects based on protein structure

[ebrintn]

For this, the idea is to change the fitness function so that instead of getting the multiplicative fitness across sites (using the fitness profiles that were selected) we evaluate whole molecule fitness by computing how 'compatible' the sequence is with a given protein structure. For this we can use something simple like Miyazawa-Jernigan contact potentials, which summarize who much particular amino-acids like to be next to each other in three dimensional space.

[ebrintn]

Calculates fitness by adding 1 and then multiplying the fitnesses

[ebrintn]

Grace has basically finished - will close issue later this week. Just need to verify with larger protein structure

[ebrintn]

Working on merging C with eidos code, should be finished by Friday, July 16

[ebrintn]

Fully functional and quite efficient - I would say it might be even more efficient than the site-heterogeneous profiles. Memory leaks have been resolved