Find a better way to dedup bigtable map proto key value pairs

Currently using the serialized string of the key proto. Since we're comparing fields of a message and not between different messages and we know they will always be serialized with the same version of protobuf, it lessens the likelihood of incorrect results. It would be better to find an officially "blessed" method of doing this.

Start by locating the Bigtable map proto key-value deduplication code and the current use of serialized key-proto strings. Investigate an officially supported protobuf comparison or equality method for fields within the same message type. Done means the deduplication approach is agreed upon and no longer relies on serialized strings.

Written by the indexing model from the issue text.