biology/paml: Update to version 4.9i

Upstream changes as described by the author:

• mcmctree: I have added an option (duplication = 1) for dating a tree with

both speciations and gene duplications, so that some nodes on the tree
share divergence times. Nodes sharing ages are identified using labels in
the tree file: #1, #2, .... I have yet to update the document about
specification of the model.

• mcmctree: The TipDate option was written for one locus or partition and

never worked for more than two loci/partitions. I have edited the code so
that it works for multiple partitions, some of which may be molecular and
the others morphological.

• codeml: The option estFreq = 0 when codonFreq = 6 (FMutSel0) and 7

(FMutSel) is not working in versions 4.9g and 4.9h. This is fixed now. This
option uses the observed codon or amino acid frequencies for the
mutation-selection models of codon usage. Instead the program estimates the
frequencies using maximum likeihood, which is what the option estFreq = 1
does. Look at the README file in the examples/mtCDNAape/ folder.

• codeml clade model D: The bounds for the w (dN/dS) ratios in the first

site classes are set tp (0.0001, 0.5) for w0 and (0.5, 1.5) for w1, in
versions 4.9b,c,d,e,f,g, since I added the BEB calculation for clade model
D in 4.9b. The motivation for the bounds is that site class 0 represents
strong purifying selection with a small w0, while site class 1 should
include sites under weak purifying selection with a larger w1. However the
bounds are arbitrary. In some datasets, the MLEs are found to be at the
bounds, making the interpretation awkward. I have changed the bounds to the
following: w0b[]={0.0001, 1.0}, w1b[]={0.01, 1.5}. This means that the user
should swap the estimates of w0 and w2 if w0 > w1.