Increase the precision of affinity maps

[pslacerda]

Albeit don't implement full precision, it may fix #392 because I don't imagine needing any more than this and is pretty much unlikely the precision set here cause any defect or slowness.

[rwxayheee]

Hi,

Can you show a benchmark (example results) of how this improves the precision of the grid maps (and how it affects the docking results)?

Also can you make an estimate, if this PR changes the size of the map files?

When I was thinking about the limited precision of using maps instead of receptor, I was more concerned about the fact that maps have limited data points, with a certain spacing, over the volume of interest. While if the interaction is computed directly, the coordinates used for calculation are true atom positions, not restricted onto the grid points.

[diogomart]

The deviation between trilinear interpolation and the true potential can be much larger than 0.001, so I second @rwxayheee question, what is the improvement here?

[pslacerda]

It was me trying to understand the code, this is not relevant anymore. Sorry for the noise.