Parallelise algorithms

This replaces !63 (closed)

• Remove all OMP #pragma statements
• Make IIntegrator::integrate and all associated code thread safe
• Add ParallelFor.h for simple loop parallelisation
• Parallelise integration, prediction, shape assignment
• Add IntegratorFactory to avoid persistent IIntegrator objects that might leak memory
• Store all images in data set in the memory by default to avoid HDF5 mishaps
• Figure of merit graphs now have d on the x-axis instead of shell number

Resolves #71 (closed), #353 (closed), #664 (closed)