Which files should be used when smashing by species?

### Context

I'm currently working on #909, which is forcing me to review/rethink some of the ways we create, associate, and use files output by the salmon processor. One thing I've realized is that since a sample can belong to more than one experiment, it may have multiple smashable files which are each a slice of a different output from tximport. Currently our smash logic just uses the most recently created file. After discussing with @cansav09 we've identified that for experiment-level smash jobs we should be making sure to use files that came from the same tximport run. However we weren't able to figure out what to do when the smash job is at the species-level.

### Problem or idea

We need to figure out a good rule for picking a computed file to use for a given sample when smashing at the species level.

### Solution or next step

Unsure, but I imagine figuring out the answer to this question will involve @jaclyn-taroni

### New Issue Checklist

- [x] The title is short and descriptive.
- [x] You have explained the context that led you to write this issue.
- [x] You have reported a problem or idea.
- [x] You have proposed a solution or next step.

Start by reviewing issue #909 and the current salmon processor smash logic. Trace how experiment-level jobs select files from a tximport run, then determine and document a consistent rule for selecting a sample's computed file during species-level smashing; the rule should resolve which file is used when multiple smashable files exist.

Written by the indexing model from the issue text.