updated params for selection, fixed bugs, now computing stats with diploshic

[mufernando]

A bunch of things:

• Changed the call ts.write_vcf, which returns 0-based positions. This is a problem bc diploSHIC assumes positions are 1-based.
• Updated the parameters of the sweeps according to what we decided on a call.
• Some refactoring to compute diploSHIC fvecs for both populations (in the sweeps simulations we introduce a sweep in each population individually -- so we have two sweep sims per param combination; however, for neutral and BGS sims, we can compute the stats in the two populations independently from the same simulation).
• Added rules to merge the stats (it's a bit ugly bc of memory usage...)
• Doubled the physical scale over which SHIC computes the stats. Right now we have CLR being computed along a grid with 20 points (about once every 250Kb), but SHIC was being computed on 100Kb windows. I was having some issues because in the simulations with sweeps variation was being completely wiped out (no snps) and SHIC simply skips windows with no SNPs.

[andrewkern]

this all LGTM @mufernando -- are there draft result plots you could post here before we merge?

[mufernando]

I just finished running all the sims and stats, so I didn't have time to run the analysis. I think we should merge this and I will open a new PR with analyses modifications and results.

[andrewkern]

okay sounds good. merging.