Bug fixes:
Fix bug in handling fixation removal (1323)
* In effect, flags to remove fixations were ignored.
This likely occurred when modifying the back end
to support demes graphs, but I have not bothered to
bisect out the source of the error.
* Added explicit C++ and Python tests of removing fixations
* Updated Python tests of selective sweeps, including
fixing some logic errors in the tests that led them
to pass "by luck" rather than by design.
This bug does not affect simulation results for multiplicative
models. For such models, relative fitness is what matters
and not removing fixations still maintains relative fitness.