scale-up

[galsalomon66]

adding the capability to execute the same query across different input-stream(CSV in this case), merge results of each of the streams, and return them to the caller as a single one. the different execution flows can run in parallel to each other

1. adding s3select_result (replacing std::string) to handle more options for result production
2. adding shared-queue to handle results of multiple and parallel execution flows
3. there are 2 main flows in query execution, 1) non-aggregate flow and 2) aggregate flow
   3a. the non-aggregate flow is mainly about merging results of different execution flows.
   3b. the aggregate flow handles the complexities of aggregation queries(sum, min, count ...), it splits the execution into 2 phases, the first is processing the query, the second is merging the results of all processes, in the case of aggregation query, it means that AST nodes, behave differently depends on thier phase.
4. s3select_scaleup simulates a multi-threaded execution of a single query. the application defines a list of files as a single input data set for the query, each of the input files executed on a dedicated thread, and a single consumer merged the results of all producers. (this flow is for the sake of simulation and measurements)
5. as for the RGW execution, multiple requests will process a single query, from the user perspective it's a single request.
6. as for splitting the input (improve scalability in some use-cases), per each data source (CSV, Parquet, JSON) it needs a different flow of input splitting.
7. TODO long result-rows should split into several entries in the shared queue.
8. TODO result should be handled by a callback or return to the caller (as value)

Signed-off-by: gal Salomon [email protected]