bigtable-client-core currently has a feature to use ReadRows for bulk reads. I think we can easily port over this functionality using the Batcher api similar to how bulk mutations work:

• Add Batcher<ByteString, Row> newBulkReadBatcher(tableId, filter) to the client and stub surface
  • Have it return a instance BatcherImpl instance that wraps the ReadRowsCallable
• Add a BulkReadRowsRequestBuilder that wraps a Query and adds ByteString row keys
• Add a ReadRowsBatchingDescriptor that can map results back to entry futures
• Expose batching read settings as bulkReadRowsSettings on the stub

Things to consider:

• Need to be careful with tracing to avoid creating confusing spans
• Need to make sure to expose the ability to set a filter
• Need to be careful with UnaryCallableSettings to make sure that bulk & streaming settings dont conflict