Fixup tests before v3

[jakobnissen]

Two things I'd like to improve on

• Some of my most recent changes here on version 2 have had insufficient testing. Ideally, I'd like to get coverage up near 100% again. This is much more easily achieved AFTER Spring cleanup for v3 #138 , so this might be one of the later issues to address before v3
• There are way too many tests for BioSequences - nearly 9000 for LongSequence and 26000 for Mers. That's in package with 10k lines! On my (rather fast) laptop, they take 1.5 minutes. Not an eternity, but still a bit of wait. Worse, most of them don't really do anything except burn CI time and spam up the terminal if something goes wrong. I bet at least 75% of these tests could be eliminated if we cut down on test patterns, like

    for len in 1:64, _ in 1:10
        @test kmer_stuff(len, rand_seq(len))
    end

Do we really need 640 distinct tests here? What are the odds that function fails at k=54, and not before?

So, plan of action:

• Complete Spring cleanup for v3 #138 , removing everything we don't want in this package
• Complete Remove Kmers #121 , and any other new features to mers or longsequences
• Aggressively cut down on needless tests for mers and longsequences
• Add tests back to get near-100% coverage

Using Aqua on Julia exposes a few method ambiguities. In general, BioSequences have an issue with method ambiguities.

• Fix method ambiguities (after all v3 feature issues, the possible interface rewrite, and Implement broadcasting for biosequences #135 )

[kescobo]

This makes sense to me as well, and is the kind of thing that I and others can productively help with. Once you've got the cleanup done, I can take a stab at (3) and/or (4).

[jakobnissen]

Since broadcasting is going to be delayed to after the v3 release, and #203 is merged, this is done.