Allow for noise correlation

Proposed by @jmartino.

We currently generate all noises using different seeds, generated as a tree of independent sub-seeds from a master (see !187 (merged)). For example, the master seed is used to create a sub-seed for the OMS noise in MOSA 12 using the key oms_12, which is then used to create the necessary sub-seeds for each noise component (requiring their own random number generator). As a consequence, all noises are uncorrelated.

We might want to study the impact of correlated, partially correlated, or anti-correlated noises. This is especially true for correlations between the MOSAs or SC of the same noise; but there might be a use case for cross-talks (correlations between different noise types).