Resolve "Speed up MWA primarybeam calculation"

This merge request aims to improve the performance of the MWAGrid::CalculatedIntegratedResponse by overriding the MakeIntegratedSnapshot by an MWA-tailored version. In particular, assuming the MWA stations are homogeneous, we only need to compute the auto-correlation of one single station for each individual pixel, rather than looping over all the baselines.

[EDIT] In addition, -O3 is enabled by default in the CXX_FLAGS. This at least results in a tremendous speed-up of the TileBeam2016::ArrayResponse method for MWA, compared to an "unoptimized" compilation.

Closes #50 (closed)