DOC: Document how weights work

[bocklund]

In ESPEI, each data point is assumed to have a residual that follows a Gaussian distribution centered at zero, μ=0, with some user-specifiable standard deviation, σ:

for different types of contributions, i. Each type of data has a default standard deviation, σ_i, which are given in the table below for the each type of data currently available. Users may provide weights, w_i, which are the product of the "weight" key in each dataset and the value for the data type in the mcmc.data_weights dictionary from the ESPEI input. The default dataset and data weights are unity, so if no weights are provided, σ_i is the standard deviation.

By controlling the weights, the standard deviation each data point can be set.

Likelihood contribution	σ_i
ZPF (expressed as driving forces)	1000 J/mol
HM	500 J/mol
SM	0.2 J/mol-K
CPM	0.2 J/mol-K
ACR (activity, expressed as chemical potentials)	500 J/mol