Previously, if a request against the default reviewers API tried to

unset one of the repositories, users, or groups fields, it would

result in an error. This was due to the API code not handling the case

of an empty string being submitted (which would get split into the list

['']) and result in us not being able to lookup the appropriate model

instances. We now filter out empty results from the list to ensure we

query the database with valid repository, user, and group names.
This change also cleans up the compare_item method for default

reviewers tests so that it compares against the provided values instead

of hardcoded ones.

Ran unit tests.