Merge duplicate indexed tool call deltas

[parsa-faraji]

Fixes #3203.

This updates the streaming delta accumulators to route initial list values through the same index-aware merge path used for later chunks. That prevents duplicate tool_calls entries with the same index in the first chunk from being stored as separate list entries.

The chat completion initial snapshot path now also normalizes choice.delta through accumulate_delta, so the first streamed chunk gets the same merge behavior as subsequent chunks.

Tests added:

• duplicate indexed list entries with both missing and None accumulator values
• primitive list deltas with both missing and None accumulator values
• chat stream state first-chunk duplicate tool call regression

Verification:

• uv run --with pytest --with pytest-asyncio python -m pytest -o addopts='' tests/lib/test_streaming_deltas.py
• uv run --with ruff ruff check src/openai/lib/streaming/_deltas.py src/openai/lib/streaming/_assistants.py src/openai/lib/streaming/chat/_completions.py tests/lib/test_streaming_deltas.py

[chatgpt-codex-connector]

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: 421986a73b

ℹ️ About Codex in GitHub

Codex has been enabled to automatically review pull requests in this repo. Reviews are triggered when you

• Open a pull request for review
• Mark a draft as ready
• Comment "@codex review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

When you sign up for Codex through ChatGPT, Codex can also answer questions or update the PR, like "@codex address that feedback".

[chatgpt-codex-connector]

Normalize first chunks for choices added later

This only applies the duplicate-index merge to choices present in the stream's very first chunk. If a multi-choice stream first sees a choice through the except IndexError path in _accumulate_chunk (for example, the first SSE has choice 0 and a later SSE first introduces choice 1) and that choice's first delta contains duplicate tool_calls entries with the same index, line 418 still stores choice.delta.to_dict() directly, so the duplicate entries remain and subsequent chunks merge into only one of them. Please route that new-choice path through accumulate_delta({}, choice.delta.to_dict()) as well.

Useful? React with 👍 / 👎.