No semantic parsing or deduplication

[jhoffmanmixmax]

Currently, the library allows for things like cheese and -cheese to be inputted, and doesn't reason about the semantic meaning of having a property and its negation both present. There's also no duplicate checking.

If we were to add duplicate checking, would we want to perform that automatically in an opinionated way? Or should it be user-configured?

How do we want to handle user input vs programmatic input? What kind of errors can we surface to the calling context which can help make the UX feel right?

example of where this has caused issues in the past:
https://github.com/mixmaxhq/app/pull/7505/files#diff-d32a3e6004a682c216ff83320b43464dR234 (private to Mixmaxers)