filtering of unused mutation types should be backed out

[bhaller]

From Jeffery Spence in Slack:

...it seems like mutationTypes with a proportion of 0 in a genomicElementType don't get added to its mutationType property, but some of the DFE testing in stdpopsim assumes that they do get added.  Did this change at some point?

An example has the following code:

initializeMutationType(1, 0.5, 'f', 0.0);
initializeMutationType(2, 0.5, 'f', 0.0);
initializeMutationType(3, 0.5, 'f', 0.0);
initializeGenomicElementType(1, c(1, 2, 3), c(0.3, 0.7, 0.));

and then running g1.mutationTypes in the terminal results in MutationType MutationType

It was agreed that the filtering out of unused mutation types should be backed out. As far as I can recall, there was no strong reason driving that change, so I think it should be an easy revert. (Famous last words.)