[Spark] Add DSv2 ReplaceData write path

[murali-db]

What changes were proposed in this pull request?

Add the copy-on-write ReplaceData write path to the Kernel-based DSv2 Spark connector, so row-level operations that rewrite touched files (e.g. UPDATE / DELETE planned as ReplaceData) can commit through the connector.

• Introduce the ReplaceData write classes:
  • DeltaReplaceDataOperation captures the configured SparkScan and exposes the scan-selected AddFiles that need to be replaced.
  • DeltaReplaceDataWriteBuilder / DeltaReplaceDataWriterFactory / DeltaReplaceDataWriter build and run the executor-side write of the replacement data files.
  • DeltaReplaceDataBatchWrite drives the Kernel transaction on the driver: it snapshots the scan-selected files up front (so commit() and createBatchWriterFactory() agree on the exact same file set), writes the new data files, and commits remove-file actions for the replaced files alongside the new add-file actions.
  • DeltaReplaceDataCommitMessage carries the per-task writer results back to the driver for commit.
• Route ReplaceData through DeltaV2WriteBuilder, capturing the SparkScan selection before newWriteBuilder() is invoked.
• Derive remove-file partition values from the scan snapshot via PartitionUtils, reusing the existing partition / _metadata plumbing in SparkScan and DeltaV2Table.
• Make the engine-info helper on DeltaV2BatchWrite package-private so the new write path can reuse it.

How was this patch tested?

• Added DeltaReplaceDataBatchWriteTest covering the driver-side transaction wiring: scan-selection snapshotting, remove-file derivation from the captured scan, and commit-message aggregation.
• Added DeltaReplaceDataWriterTest covering the executor-side writer factory and writer, including the data-file commit messages they produce.

Does this PR introduce any user-facing change?

No. The DSv2 Kernel connector lives under io.delta.spark.internal.v2; this change adds an internal write path and does not change any public API or existing behavior.