MRG: update to (prospective) sourmash r0.18.0

[ctb]

This PR updates to sourmash core latest, which will be r0.18.0.

The biggest set of changes in the plugin relate to sourmash-bio/sourmash#3431, which made it possible to detect bad (and good) zip files based on content, and handle the error; in particular, this means that we can continue past errors from reading a file misnamed as a sig zip file, and try reading it in other ways. So a bunch of tests (e.g. test_fastgather.py::test_bad_against_3) no longer apply!

sourmash r0.18.0 also includes sourmash-bio/sourmash#3434, which changes the behavior of Signature::name() to return an Option if name is not set, rather than defaulting to either filename() or md5sum(), in that order. This caused a bug (see #538 (comment), split out into #550) where manifests would improperly disagree about whether two Records matched.

Fixes #550

TODO:

• add test for manifest intersection per intersect_manifest fails to match identical Records because of Signature::name() behavior #550
• should we add back a test for a misnamed zip file?

[ctb]

Ready for review & merge @bluegenes! But I will merge this w/o a review as needed in the interest of getting #531 into a release ;).

[bluegenes]

Lgtm

[bluegenes]

We still have a test for improper zip file that can't be read, right?

[ctb]

Uhh... actually, good question. Probably. What kind(s) of things are you thinking about? I could see -

• zip file that is corrupted
• zip file that doesn't have a manifest

other things?

[bluegenes]

Yep -- I was mostly thinking about zipfiles that get interrupted while writing and thus can't be read - I guess this falls into your "corrupted" category. The error is now propagated from sourmash, so I'm guessing we just exit and say we can't load the sigs.

Just wanted to make sure we still have a test covering this.