Where duplicate index documents exist, the result count is indeterminate

[p-kimberley]

In an index where multiple documents exist with the same StreamId and EventId, search results will vary between searches using the exact same expression tree.

To the user, this manifests as a varying result count.

Ideally no such duplication should exist. However in the event it does, duplicates should be filtered out consistently, in a deterministic manner.