Fix misplaced indel edits

If you map read S1_73477 from the 1m simulated R10 reads with vg ef2a1384b, like this:

READ_NAME=S1_73477
vg filter --exact-name -n "${READ_NAME}" /private/groups/patenlab/anovak/projects/hprc/lr-giraffe/reads/sim/r10/HG002/HG002-sim-r10-1m.gam >read.truth.gam
GRAPH_BASE=/private/groups/patenlab/anovak/projects/hprc/lr-giraffe/graphs/hprc-v1.1-mc-chm13.d9
MINPARAMS=k31.w50.W
vg giraffe -t16 --parameter-preset r10 --track-provenance --track-correctness --progress --show-work -G read.truth.gam -Z ${GRAPH_BASE}.gbz -d ${GRAPH_BASE}.dist -m ${GRAPH_BASE}.${MINPARAMS}.withzip.min -z ${GRAPH_BASE}.${MINPARAMS}.zipcodes -x ${GRAPH_BASE}.xg >remapped.gam 2>log-${READ_NAME}.txt

Then the alignment doesn't validate:

vg validate -A -a remapped.gam /private/groups/patenlab/anovak/projects/hprc/lr-giraffe/graphs/hprc-v1.1-mc-chm13.d9.gbz

Invalid Alignment:

Length of node 4465278 (1) exceeded by Mapping with offset 0 and from-length 4:
{"edit": [{"from_length": 1, "to_length": 1}, {"sequence": "AAGG", "to_length": 4}, {"from_length": 3}], "position": {"node_id": "4465278"}, "rank": "63"}
alignment: invalid

This is because a "from_length": 3 deletion edit that might make sense in a mapping to node 4465277 is smushed into the mapping for the previous node, node 4465278. It also happens to be immediately after an insertion edit; generally we don't want to abut those.

We have to figure out how this base-level alignment is being generated, and at least get the edit into the right mapping, if not prohibit the adjacent indels entirely.

Without this, surject can't process the read.

vgteam / vg

Fix misplaced indel edits #4320