finalise the selection API

[grahamgower]

#462 added support for doing single locus selection in a flexible way, and DFEs in an inflexible way. Probably the API for using this will need to change as we demand more, or more complex, features. @jeromekelleher has expressed concern over the location of this code (it currently lives in the ext submodule, which is unecessary), and the ExtendedEvents objects.

[mufernando]

why do you think the DFE API is inflexible?

[grahamgower]

Possibly just a poor choice of words. :) Combining the MutationType class with a new GenomicElementType class would certainly make it more flexible though.

[izabelcavassim]

I agree with @jeromekelleher that the ext submodule is unnecessary, maybe we could fix this @grahamgower and @petrelharp, before we push the DFE_class PR?

[petrelharp]

Let's do it in a separate PR.

[nspope]

Closed by #1341. Selection-related functions are still living in ext, I'll make a new issue to track moving these somewhere more appropriate.