[VL] Add E2E spill-to-disk benchmark

Should add some E2E benchmark code measuring on Velox's spill-to-disk performance in Gluten's code base. For more precise reflection to production use case, it's preferred to add the benchmark code as a Spark benchmark suite.

Start by locating the existing Spark benchmark suites in the Gluten code base and reviewing how they model production-like workloads. Use the Spark benchmark suite entry point to determine how Velox spill-to-disk execution is measured; done means an E2E benchmark reports that performance clearly.

Written by the indexing model from the issue text.