[core] Support row-count based bucket calculation for postpone compact.

[Stephen0421]

Purpose

Support calculating postpone bucket count by configured target row count per bucket.

A new table option is added:

• postpone.target-row-num-per-bucket

When compacting postpone bucket files, bucket count is resolved in this order:

1. Reuse known bucket count if the partition was already compacted before.
2. If postpone.target-row-num-per-bucket is configured, compute bucket count from active postpone row count.
3. Fall back to postpone.default-bucket-num.

Changes

• Added CoreOptions.POSTPONE_TARGET_ROW_NUM_PER_BUCKET.
• Added shared PostponeUtils.determineBucketNum to keep Flink and Spark behavior consistent.
• Added row-count based bucket calculation helper and tests.
• Updated Flink postpone compact action to use the shared bucket decision logic.
• Updated Spark postpone compact procedure to avoid capturing java.util.Optional in Spark task closures.
• Improved overflow error message with user guidance.
• Updated postpone bucket documentation.

Test

• org.apache.paimon.flink.PostponeBucketTableITCase
• org.apache.paimon.spark.sql.PostponeBucketTableTest
• org.apache.paimon.table.PostponeUtilsTest