augur ancestral so far only dealt with nucleotide sequences, while
translation were done codon by codon from the nucleotide sequences in
later steps. If translations for tips are available, the ancestral
amino acid sequences can be determined analogously via ancestral
reconstruction. This is implemented in this commit.

For each node, the node data structure will contain a `muts` and a
`aa_muts` field. The latter is a dict with gene/cds names as keys and
`<ancestral><pos><derived>` mutations (same as in augur translate).
In addition, an annotation and reference sequences are derived stored
in the node-data-json.

…notation and translation

Reorganize functional tests for augur ancestral to follow the standard
pattern with "cram" and "data" subdirectories and separate cram files
for individual functional tests. Paves the way for adding another
functional test for amino acid sequence reconstruction.

Adds amino acid sequences for the corresponding nucleotide alignment in
the ancestral data directory (created from augur translate but mimicking
output from Nextalign) and adds a functional test for the new interface
that allows reconstruction of amino acid sequences for internal nodes
from the given tip sequences per gene.