Feature/corr cholesky [WIP]

[adamhaber]

This PR implements two different ways to transform a vector of K(K-1)/2 real numbers to a Cholesky factor of a K x K correlation matrix, addressing #24:

1. Stan's way, based on https://github.com/stan-dev/math/blob/92075708b1d1796eb82e3b284cd11e544433518e/stan/math/rev/fun/cholesky_corr_constrain.hpp#L38-L52

2. TFP's way, based on https://github.com/tensorflow/probability/blob/v0.17.0/tensorflow_probability/python/bijectors/correlation_cholesky.py

I just want to make sure this is a sensible first step in the proposed workflow; I'll then address the other points raised in #24.

One thing I'm not sure about in the Stan implementation: as far as I understand, in this line we are accounting for the jacobian correction of the tanh transform, whereas in this line we do not; why is that? And, presumably, here we do want to account for that, right? (currently we do not, I need to change that).

[bob-carpenter]

Some of these are stylistic comments you can ignore. But it should have a little doc indicating which variable gets a uniform distribution over constrained space and which variable is the unconstrained parameter.

[adamhaber]

Thanks for the review. If this seems alright, I think the reasonable next steps would be:

1. Compare against TFP and make sure we get the same results (numerically); specifically, make sure we get the log_det_jacobian correction right.
2. Run initial evaluations.

Sounds good?

[bob-carpenter]

@admahaber: Yes, that sounds great.