Parallelise assignment functions

[johnlees]

In models.py can do chunks at a time with sharedmem

[nickjcroucher]

Possible with numba pthreads, to maintain optimisation of each process? Numba seems to be incompatible with multiprocessing at the moment. I phrase as a question, as I've been trying without anything to hold up as evidence of this being possible.

[nickjcroucher]

Functioning example for refined models: 8cb08c9.

[nickjcroucher]

Wrong commit - see 513729a. Pthreads shares read-only so no good for matrix translation.

[nickjcroucher]

Just recording a thought - in models.py, plot re-assigns sampled points - currently not multi-threaded.

[nickjcroucher]

Need to think about this with regard to numba: https://numba.pydata.org/numba-doc/latest/user/faq.html.