Inconsistency when Indexing Duplicates in Redis

[rfecher]

When an index produces duplicates for a single result, Redis only allows one entry in a collection with the same value (so there may be a different key, but the purpose of duplicates is to have the exact same value). Redis behavior is to overwrite with the most recent matching value within the collection. To maintain all backwards compatibility but allow for duplicates to behave properly, there is an optional, nillable duplicateId that is persisted in the collection (assuming that the redis writer is given an array of geowave rows with all duplicates, if this assumption is wrong, at best we could just generate a random duplicate ID if the other duplicate values are not in context at the same time).