fix: apply retriever weights in reciprocal rerank fusion

[wolfkill]

Summary

• Apply retriever_weights when calculating reciprocal rerank fusion scores
• Preserve the retriever index by iterating over result items instead of values
• Add a regression test for mode=reciprocal_rerank with a zero-weight retriever

Fixes #21444

Testing

• UV_CACHE_DIR=.uv-cache uv run --group dev pytest tests/retrievers/test_fusion_retriever.py -q
• UV_CACHE_DIR=.uv-cache uv run --group dev ruff check llama_index/core/retrievers/fusion_retriever.py tests/retrievers/test_fusion_retriever.py
• UV_CACHE_DIR=.uv-cache uv run --group dev ruff format --check llama_index/core/retrievers/fusion_retriever.py tests/retrievers/test_fusion_retriever.py

[gautamvarmadatla]

Hi @wolfkill, thank you so much for working on this. I already created a PR for this along with issue #21445, so we should be covered there.

[github-actions]

This PR is stale because it has been open 50 days with no activity. Remove stale label or comment or this will be closed in 10 days.

[logan-markewich]

This is being closed as a detected duplicate of #21445. If you feel otherwise, feel free to re-open a fresh PR