plot_pairwise_average_fst may change the order of the cohorts when using a cohort dict

[jonbrenas]

When using user defined cohorts using a dictionary, say:

plot_pairwise_average_fst (and I assume some other functions as well) will reorder them so that they are in alphabetical order. This is a problem when, e.g., the annotate_se option is used because the se and fst values are swapped, e.g. using the dictionary from the above example:

Another time when that is a problem is when the user ordered the keys for a good reason.