[flink] support combined mode for orphan files clean

[XiaoHongbo-Hope]

This PR

Supports combined mode processing when orphan files clean:

• Processes multiple tables within a single DataStream during job graph construction, instead of creating one DataStream per table, which significantly reduces JobGraph construction time and complexity, avoiding timeout, stack overflow, and resource allocation failures
• Only applies when --mode combined is specified

Adds configuration:

• [--mode <divided|combined>]: Processing mode (default: divided)
  • divided: Create one DataStream per table (original behavior)
  • combined: Process all tables in a single DataStream
• [--tables <table1>] [--tables <table2>]: multiple parameters for table names
• --table and --tables cannot be used together

Tests:

• testCombinedMode: Combined mode with multiple tables
• testCombinedModeWithBranch: Combined mode with multiple branches

[XiaoHongbo-Hope]

@yuzelin @jerry-024 @JingsongLi Can you please help review this PR.

[yuzelin]

I think it's better to use one flink datastream to handle all tables. For example, you can add a --tables a,b,c , and it means clean table a, b and c; If not specified, get tables from catalog by old style arguments.

[JingsongLi]

Is combine mode any bad case? Why not just enable it by default

[XiaoHongbo-Hope]

Is combine mode any bad case? Why not just enable it by default

Sure，we can enable it by default.

[XiaoHongbo-Hope]

Is combine mode any bad case? Why not just enable it by default

Updated and tested in our application.