Fix references to "flat" priors in docs

[nspope]

as we actually do use a "prior" (really, regularisation) on root node ages.

[hyanwong]

But we do use a flat prior on non-root ages, right?

[hyanwong]

I think it would be good to keep the idea that the prior on non-roots is essentially flat (after all, these are most of the nodes we are trying to date: the roots are an exception)

[nspope]

What we're doing is using regularisation on the roots-- fit by EM. If viewed as a prior (ignoring the EM) this induces a distribution on the internal nodes, too. I agree, it's much less strong than using informative priors on every node. But I think it's good to be clear about what we're doing here, until we write up the algorithm.