Use cohort hash and label hash in matrix and model group definitions

If a user updates the query but not the name, we should still remake the matrices and models, even if the user doesn't set the replace flag.

As is, the nice thing is that if you discover a bug, you can set the replace flag and your old models and matrices will be overwritten, simplifying your storage. However, forgetful users may neglect setting the replace flag or updating the name and encounter unexpected behavior, such as overwriting valid matrices or not building new models when it is expected that they will be built.

Of course, we probably wouldn't want to make this change until there is a good cleanup utility that would allow the erasure of matrices/models built from erroneous queries.

Start by locating the matrix and model group definitions and how they use the replace flag; the issue does not name files or tests. The proposed behavior depends on cohort and label hashes, but the issue says a cleanup utility should come first. Done means changed queries rebuild matrices and models without requiring a name change or replace flag, with a way to clean up outputs from erroneous queries.

Written by the indexing model from the issue text.