issue with data simulation under codon Markov-modulated model

[GuyBaele]

When simulating data under a codon MMM, the simulated sequences look as follows:
TTA0AAT0GTA0TCT1 ...
While this is very useful for parsing output annotated on trees, it prevents immediately using these simulated sequences in an XML for analysis, so the numbers have to be manually removed.
Nucleotide MMMs generate different output for tree annotation and simulation, and we should make sure codon MMMs do so as well, but should also make the different types of output coherent between different types of MMM.