Rates and traits confidence v2

[trvrb]

This is a more-or-less direct rebase of PR #213. It does three things:

1. Implements a --sampling-bias-correction flag in augur traits to add uncertainty to ancestral trait reconstruction.
2. Implements a --clock-std-dev flag in augur refine to allow inputting of standard deviation of "fixed" clock rate.
3. Adds both these options to the test builds.
4. Upgrades TreeTime to version 0.5.1.

---

@rneher: My testing is showing no effect from the --sampling-bias-correction flag. I tested and it does modify tt.gtr.mu correctly. However, this seems to have no impact on resulting reconstructions. Can you take a look at this?

[rneher]

For me it does have an impact on the reconstructions -- the higher the value the higher the entropy of the trait reconstruction. What exactly are you comparing?

[trvrb]

@rneher ---

I see now. I was somehow expecting more of an impact. In the standard Zika build going from --sampling-bias-correction of 1 to 3 takes the root of the tree from 48% Brazil to 43% Brazil.

I suppose that this is due in part to a sparse migration matrix so that adjusting rate only does not have a huge smoothing impact. Does this make sense?

[trvrb]

I'm going to merge this now and we can continue to improve this aspect. Everything else is working great.