Make a pipeline for only the binary classifier

[mihow]

We need a method to run only the binary classifier and show results from that. We also have a request to always return non-moth results as well (rather than filter them from the response). Currently the binary filter does not send non-moth's to the species classifier stage, that behavior can continue, but always return all results. The filtering could happen within the classifier itself.

• The ML backend should always return moth & non-moth classifications
• If a species classifier is requested, only add classifications to the moth classifications (run filter)
• It's okay if the binary classifier has to run twice to simplify the pipeline & configurability.

Other issues to look at while in there:

• Allow skipping the object detector
• Find where duplicate classifications are coming from

[mihow]

Closed by RolnickLab/ami-data-companion#70 and #684