The time series fuzzer creates a normal collection and a time series collection, and runs queries to make sure results are the same. We should take that grammar, reduce it to only stages and accumulators supported by block processing, and create a new TS-block fuzzer.