replace rcpp backend

[ajrominger]

[ajrominger]

• return roleData S4 object for every snapshot
• in engine-replacement branch, add actually tracking of harmonic mean and first arrival #156
• figure out best way to deal with scale, generations, and J when updating phylo and handing off to msprime #153
• check initialization of harmonic mean, last origin, last extinction, and prop equilib in R and make sure they are properly passed to simOO
• make trait data live in matrix to support multiple trait axes
• change class of edge and edgeLength in rcpp updatePhylo module so we don't have to cast them to different types #138
• make bigger env_sigma mean wider env kernal, meaning you need to be even closer to optim to benefit from it #154
• check that partial neutrality actually has an impact (right now only seems that any kind of neutrality turns on only for delta = 1) #155
• fix bug where trait right at optimum sometimes leads to "too few positive probabilities" in sampling death #157