Gather test data for monorepos

[rarkins]

Currently most monorepo updates such as for @angular show low data because we currently exclude branches with more than one update from our statistics. However, for packages which are nearly always upgraded together, this means we filter out all the data!

First step will be to prefer source URL for each package and base our filtering on unique source URLs instead of package names.

Second step will be to include weighted scores for multi-source PRs, especially if they pass.

[rarkins]

We should try to evaluate this against some common monorepos such as Angular