path editing yields dangling nodes when they are soft clips off the end of an existing node

[ekg]

To reproduce, from test/

vg construct -v tiny/tiny.vcf.gz -r tiny/tiny.fa >t.vg
vg msga -g t.vg -s TTCTATAATATG | vg stats -s -

This shows that there are now two disjoint subgraphs (whoops):

1       1
2       55

The original graph:

Post MSGA:

The G node corresponds to the last base of the read. It's aligning right at the end of the node and isn't being dropped in properly.

... resolving

[ekg]

It also works in the other direction, such as a soft clip at the beginning of the alignment:

vg msga -g t.vg -s CAAATTTTCTGGAGTTCTAT

So this seems to be a balanced problem.