Add toggle for treatment of MNPs in VETS.

[samuelklee]

The current implementation of VETS borrowed VQSR's logic for classifying variants into SNPs and indels, for which separate models are trained. We retained this logic to make comparisons with VQSR as straightforward as possible.

In this logic, which originates from htsjdk, an alternate allele with len(REF) = len(ALT) is counted as a MNP and classified as a SNP. However, we are now applying VETS to long-read genotyping of SVs, where inversions satisfy the same criterion but are then awkwardly classified as SNPs.

We should add a toggle to allow classification of MNPs as indels, probably retaining the old behavior as default to not cause any changes for e.g. GVS.

Tagging @koncheto-broad and @fabio-cunial for their visibility.