fix: avoid existing chunk duplication in multi-node optimize(mode="ap…

[hillhack]

What does this PR do?

Fixes #865

Fixes duplicate existing chunks when optimize(..., mode="append") is used with multiple nodes.

Previously, each node included the existing dataset chunks in its intermediate index. During the final merge, these chunks were combined multiple times, resulting in duplication.

Changes

• DataProcessor._done(): Excludes existing_index from intermediate node-level merges when using multiple nodes, while preserving the existing behavior for single-node append.
• DataProcessor._upload_index(): Adds existing_index only once during the final global merge.
• Tests: Added test_multinode_done_append_no_duplication to verify that existing chunks appear exactly once in the final index.

Expected result

For an existing dataset [A, B] and new chunks [C, D] produced by two nodes:

Before: [A, B, C, A, B, D]
After:  [A, B, C, D]