Expanding biochemical model to IID data, experiments with replicates and multiple conditons

[robertness]

This expands the biochemical model to deal with IID data. The mean of the noise distributions is a tensor of the appropriate shape, and transmitters.py includes the use of pyro.iarange to declare the dimensions independent. Pyro distributions distinguish between the "batch shape" and "event shape" of a tensor. This code assumes that "batch shape" refers to independent samples, and "event shape" is empty.

It would be preferable to use the "sample_shape" argument of the sample method of a distribution object. But it seems that passing the "sample_shape" keyword argument to pyro.sample causes and error during inference.