Generic functions for cohort means and sums

[timothymillar]

I've come across a couple of situations where I want to take mean of sample values for each cohort in a dataset. From what I can tell, there is not a simple way to achieve this using dask or xarray.

We already have a numba jitted function to do this for a specific case (_cohort_observed_heterozygosity) so it would make sense to generalize that to standard function in utils.py or similar. A higher level function could be made available which automatically calculates the cohort dimension size and runs da.mapblocks etc. We could also to the same for cohort sums.

[jeromekelleher]

Sounds great @timothymillar!