[VL] Lower default spill run size to reduce overhead memory usage

[zhztheplayer]

It's observed in CI Q67/Q95 OOM test that large global spill pool usage is causing overhead memory OOM. The PR tries to tweak on default config option of Velox spill run size to fix the issue.

The OOM message was like:

"allocateNonContiguous failed with 2 pages from Memory Pool[__sys_spilling__ LEAF root[__sys_root__] parent[__sys_root__] MALLOC no-usage-track thread-safe]<unlimited max capacity unlimited capacity used 0B available 0B reservation [used 0B, reserved 0B, min 0B] counters [allocs 149202, frees 121609, reserves 0, releases 0, collisions 0])> Exceeded memory allocator limit when allocating 2 new pages, the memory allocator capacity is 1.35GB"

[github-actions]

Thanks for opening a pull request!

Could you open an issue for this pull request on Github Issues?

https://github.com/apache/incubator-gluten/issues

Then could you also rename commit message and pull request title in the following format?

[GLUTEN-${ISSUES_ID}][COMPONENT]feat/fix: ${detailed message}

See also:

• Other pull requests

[github-actions]

Run Gluten Clickhouse CI

[FelixYBW]

why recent velox PR caused the error?

[github-actions]

Run Gluten Clickhouse CI

[zhouyuan]

These two commits maybe related:
Don't reclaim from a hash build/probe operators if it has exceeded max spill level limit (#11123)

Free hash table after grouping set/row number spill to release memory plus a hash table fix (#11180)