Insertion counts for small insertions that change the reading frame

Sample JLAT1-HIV_S217 from run 190501_M05995 has an insertion of length 2 in GP120. Because this insertion is in the middle of a codon and changes the reading frame, there is a non-covered amino acid in amino.csv at this position (at reference amino position 60, query nuc positions 5753 and 5754 are insertions).
However, no insertion counts are recorded in amino.csv and nuc.csv. This is because the code that records insertions treats them as chunks of 3 bases, and therefore misses this small insertion of length 2, see here:
The best solution would probably be to switch the insertion counting to 1-based positions to better reflect insertions whose length is not a multiple of three.

Start in micall/utils/consensus_aligner.py around the insertion-counting code linked in the issue, and trace how insertions reach amino.csv and nuc.csv. Reproduce the JLAT1-HIV_S217 example from run 190501_M05995, then verify that the length-2 insertion at reference amino position 60 is counted in both outputs without breaking other insertion cases.