[FLINK-37384][runtime] Optimize the parallelism decide logic for pointwise

[noorall]

What is the purpose of the change

When a Flink job contains a Rescale edge converted from Hash, the pre-determined parallelism limits the flexibility of parallelism derivation for Pointwise-type inputs. This consequently leads to a high likelihood of needing to use the adjustToClosestLegalParallelism method in this scenario to adjust the parallelism to meet expectations. This adjustment incurs a significant time complexity (O(log n * n * n), specific to this scenario), so we need to optimize the parallelism decision logic for Pointwise to enhance performance in this case.

Brief change log

Optimize the parallelism decide logic for pointwise.

Does this pull request potentially affect one of the following parts:

• Dependencies (does it add or upgrade a dependency): (yes / no)
• The public API, i.e., is any changed class annotated with @Public(Evolving): (yes / no)
• The serializers: (yes / no / don't know)
• The runtime per-record code paths (performance sensitive): (yes / no / don't know)
• Anything that affects deployment or recovery: JobManager (and its components), Checkpointing, Kubernetes/Yarn, ZooKeeper: (yes / no / don't know)
• The S3 file system connector: (yes / no / don't know)

Documentation

• Does this pull request introduce a new feature? (yes / no)
• If yes, how is the feature documented? (not applicable / docs / JavaDocs / not documented)

[flinkbot]

CI report:

• f534fa4 Azure: SUCCESS

Bot commands

The @flinkbot bot supports the following commands:

• @flinkbot run azure re-run the last Azure build