write brms model for zero inflated and contaminated regression

[rmcminds]

function mixture() in brms allows for the creation of a mixture model, which could allow for a 'low-value-inflated' regression (like txzhou/mixtureReg). I think a mixture(poisson, zero_inflated_poisson) would be the proper family, unless an explicit three-way mixture could be created. should be able to e.g. redo the immune allometry analysis with this brms model on a gene-by-gene basis using normalizing factors instead of a single multinomial.