Tracking issue of result cache and download

[youngsofun]

Summary

background

Originally posted by @BohuTANG in #4521

result table is used in multi places:

• downloading with formats
• table function like result_scan(<query_id>)
• internal cache for other queries. see Cache: Query result cache #4521

Roadmap

• able to download result base on result table Feat(httphandler): result download #5395
  • write to result table when serving /v1/query
  • read result table to to serve `/v1/query/<query_id>/download
  • use result table to make /v1/query not blocked by client due to limited memory buffer
• support more download format
• management of result table, delete outdated
• use table function to read it from table(result_scan(last_query_id()))
• optimize write pipeline with 2 sinks
• clickhouse/mysql write to result table too.
• use result table to optimize query
• avoid huge result Tracking issue of result cache and download #5398 (comment)
• download without bendquery，

optional：

• download limit rows，
• download parallel

[BohuTANG]

If select * from a huge table(with TB data size), we should reject the result cache.