⏱️ feat: Configurable Timeouts for the Web-Search Reranker and Image Generation

[kevinpark1217]

Summary

Two tool timeouts are fixed today, and slower self-hosted backends exceed them: the web-search rerank (10 s per page in @librechat/agents) and the image_gen_oai OpenAI client. This PR makes both configurable:

• webSearch.rerankerTimeout (ms), passed to createSearchTool, which already accepts it.
• IMAGE_GEN_OAI_TIMEOUT_MS and IMAGE_GEN_OAI_MAX_RETRIES, passed to the image_gen_oai client as timeout and maxRetries. These are env vars because the tool's other settings are too.

The image timeout is also passed to the request's undici dispatcher as headersTimeout/bodyTimeout, the same way the chat endpoints use transportTimeouts. Without it, undici's 300 s defaults end the request first, and openai-node reports that as Request timed out.

When unset, behavior is unchanged.

Related to #16494
Related to LibreChat-AI/agents#290

Type of change

• Feature

Testing

Added tests for schema validation, env parsing, dispatcher selection (direct and proxy), the authResult passthrough, and the options the OpenAI client receives. I also checked both settings against stub endpoints. With rerankerTimeout: 500, a 1.5 s rerank is aborted at ~500 ms. With IMAGE_GEN_OAI_TIMEOUT_MS=1800000, an image response held for 330 s succeeds; with only openai's timeout set, the same request fails at ~300 s.

Screenshots / recordings

No user-facing change.

Checklist

• I reviewed my own changes
• Relevant tests have been added or updated
• Existing relevant tests pass
• The change does not introduce new warnings or errors
• User-facing or complex behavior is documented where necessary
• Required dependency changes have been merged/published
• Required documentation PR: N/A