Timing of genetic value calculation vs fixation removal

[molpopgen]

PR #1324 fixed an issue where fixations were not being "pruned" even when that option was True.

With that fix, though, we have the following behaviors:

• genetic values are calculated BEFORE and fixation pruning
• Therefore, fixations are part of the genetic value calculation!!
  Including them is the right thing to do for additive model and is
  "fine" for multiplicative models b/c the way we choose parents
  only requires that we know relative fitnesses.
• However, certain operations may be annoying.
  For example, tracking mean fitness over time in "standard pop gen" scenarios
  requires correcting values stored in metadata by the effect sizes
  of fixations that generation.
  (This situation is even more obnoxious when we are not tracking
  counts each generation -- fixations since the last simplification
  accumulate.)

Ultimately, these issues are all due to fwpdy11 being written to simulate
traits and all the "pop gen" stuff is bolted on top.

It may be possible to move the order of when genetic values are calculated
to AFTER fixation removal.
Dealing with the simplification issue described above is harder -- we don't necessarily
know the counts.

[molpopgen]

Quick experimentation changing the order of events results in most tests passing but the sweep tests seem to loop forever, implying that how we are monitoring for fixations is depending on the mutations being found in genomes.