Introduce a new sink distribution mode `BUCKET_LOAD_BALANCE` for the Flink connector.

[loserwang1024]

Search before asking

• I searched in the issues and found nothing similar.

Motivation

For tables with bucket keys, BUCKET mode provides good locality and ordering guarantees, but
it can suffer from severe data skew when the number of buckets is not evenly divisible by the
sink parallelism, or when some buckets are hotter than others. NONE mode avoids skew but
breaks merge-engine semantics because records with the same key may be processed by different
subtasks.

Unlike the existing BUCKET mode (which maps each bucket to exactly one subtask and may leave
subtasks idle), BUCKET_LOAD_BALANCE distributes records by bucket key using an LCM-based
logical slot assignment so that every subtask receives traffic while still guaranteeing that
records with the same bucket key route to the same subtask.

Solution

• Records with the same bucket key always route to the same subtask.
• All subtasks receive data when numBuckets and numChannels are not evenly divisible.
• Intra-bucket ordering is not preserved (records with different bucket keys in the same
  bucket may be processed by different subtasks).

Anything else?

No response

Willingness to contribute

• I'm willing to submit a PR!