equal treatment for supplementary alignments

[tedsharpe]

Allow no sequence in supplementary alignments, just as we do for secondary alignments.

[lbergelson]

Looks good. How was this problem manifesting? Bad split guesses? Is there a reasonable test we could add to show that it's fixed now?

[tedsharpe]

Problem manifests when supplementary alignments have no sequence: guesser doesn't recognize a valid record start. This change just parallels an earlier commit that allowed no-sequence secondary alignments, so that supplementaries now behave the same way.
Tweaked data to demonstrate validity of fix.

[lbergelson]

I think you accidentally included the snapshot version change here. Could you revert that?

[tedsharpe]

Sure, but I thought we'd want to bump the version with this PR.

[lbergelson]

It gets bumped automatically as part of the release process. If we bump it manually we'll end up skipping a version.

[lbergelson]

@tedsharpe Thanks for amending the test data. What happens without this change? Does it break the split guessing so it eventually runs off the end of the file without find a split? Or a crash, or a bad split that causes data corruption? I'm just curious.

[tedsharpe]

Assuming no splitting index: If the density of supplementaries without sequence is high enough that you never get a run of 10 consecutive primary lines in some partition, the partition will be empty.
It's a weird edge case, but the bams I make from local assemblies in SV regions are queryname sorted, and have a lot of supplementaries. That triggers this ugliness.

[lbergelson]

just bump back the version and good to merge

[heuermh]

Thank you, @tedsharpe!