Accelerate TPC-H without any fallbacks to Spark

[andygrove]

What is the problem the feature request solves?

Comet now accelerates most of TPC-H, but there are still a small number of fallbacks happening.

q11/20:

WARN CometSparkSessionExtensions: Comet cannot accelerate Cast because: cast(ps_availqty#343 as decimal(10,0)) is not fully compatible with Spark (No overflow check)

q11:

WARN CometSparkSessionExtensions: Comet cannot accelerate LocalTableScanExec because: LocalTableScan is not supported

q16:

WARN CometSparkSessionExtensions: Comet cannot accelerate BroadcastHashJoinExec because: BuildRight with LeftAnti is not supported

q21:

WARN CometSparkSessionExtensions: Comet cannot accelerate SortMergeJoinExec because: spark.comet.exec.sortMergeJoinWithJoinFilter.enabled is not enabled

Issues:

• Add support for native hash join with LeftAnti + BuildRight #1909
• Implement Spark-compatible cast from integral types to decimal #2049
• Performance regression after adding support for SMJ with join filter #901
• Support LocalTableScanExec #2513

Describe the potential solution

No response

Additional context

No response