dbeta_ds: Derivative of betabinomial density w.r.t. bias parameter.
In dcgerard/updogAlpha: Using Parental Data for Offspring Genotyping

Uses chain rule with dbeta_dprop * dxi_dd * dd_ds.

1	dbeta_ds(x, n, s, ell, p, h)

`x`	The number of counts of reference allele.
`n`	The number of counts of reads.
`s`	We have `s = exp(d)`, where `d` is the bias parameter.
`ell`	Expit of eps. We have eps = exp(ell) / (1 + exp(ell)), or ell = log(eps / (1 - eps))
`p`	The proportion of genome that is the reference allele.
`h`	the overdisperion parameter. tau = 1 / (h + 1) or h = (1 - tau) / tau