Make random anoph testing data reproducible

The simulated data is generated randomly, but it needs to be reproducible
after the fact for debugging if there is a failure in one of the test cases.
This is done by generating a different global seed on each run, and then
creating context-specific RNG's in every situation they are needed. The context
is mixed into the seed of the RNG to ensure that all contexts will generate
independent random numbers, but still be reproducible from the global seed.
See https://numpy.org/doc/stable/reference/random/parallel.html#sequence-of-integer-seeds
In order for reproducibility to work, it is essential to *not* use the
python random library or any legacy numpy.random calls to generate data.

Author: Jacob Garber