[FEATURE]: Add provider timeouts and jittered retry handling for LLM API calls

[michaelmanly]

Feature hasn't been suggested before.

• I have verified this feature I'm about to request hasn't been suggested before.

Describe the enhancement you want to request

Summary

This PR improves production reliability for LLM provider calls by adding safer timeout and retry behavior around OpenAI, Anthropic, and Gemini requests.

Problem

A few provider paths currently appear to depend on caller-provided context behavior and provider-specific retry handling:

• OpenAI, Anthropic, and Gemini calls pass ctx directly into SDK calls without applying a provider-layer timeout
• title generation uses context.Background(), so it can continue after the parent request is cancelled
• retry backoff appears deterministic rather than jittered
• retry classification differs across providers and misses some transient failures
• Gemini streaming retry behavior can enter an inconsistent state after stream errors
• Gemini client init can return a provider with a nil client after logging an error
• some Gemini tool-call JSON parse/marshal errors are ignored

Changes

• Add provider-layer timeout/deadline handling around LLM API calls
• Derive title-generation context from the parent request with a short timeout
• Replace deterministic retry delay with jittered backoff
• Centralize retry classification for transient provider failures
• Include common retryable cases such as:
  • 429
  • 500
  • 502
  • 503
  • 504
  • transport resets/timeouts where detectable
• Ensure Gemini stream retry exits and restarts cleanly instead of continuing in a partial stream state
• Return Gemini client initialization errors instead of creating a provider with a nil client
• Add structured warnings for ignored Gemini tool-call JSON conversion errors

Why this matters

These changes reduce the chance of:

• hung provider calls tying up goroutines/session slots
• orphaned title-generation requests continuing after cancellation
• synchronized retry storms during provider 429/5xx incidents
• avoidable user-visible failures during short provider outages
• duplicate/truncated Gemini streaming output after retryable stream errors
• nil-client runtime failures after Gemini initialization errors
• silent tool-call degradation from ignored JSON conversion failures

Notes

This PR does not add provider/model fallback routing yet. That can be handled separately as a follow-up because it needs policy decisions around cost, model compatibility, and output behavior.

[github-actions]

This issue doesn't fully meet our contributing guidelines.

What needs to be fixed:

• The description reads like a pull request rather than a feature request — it uses PR language ("This PR improves...", "## Changes", "## Notes") and describes implementation details rather than the user-facing problem or need. Feature requests should explain the problem you're experiencing and the desired behavior, not a proposed implementation diff.
• The content appears AI-generated (highly structured, PR-style format). Please describe the feature in your own words, focusing on what problem you want solved and why it matters to you.

Please edit this issue to address the above within 2 hours, or it will be automatically closed.

---

This issue might be a duplicate of existing issues. Please check:

• LLM provider SDK lacks retry logic for transient API errors #11705: LLM provider SDK lacks retry logic for transient API errors (same request: retry backoff, jitter, 429/5xx handling)
• [Feature Request] Configurable Retry Mechanism via opencode.json #3011: Configurable Retry Mechanism via opencode.json (same request: configurable retry behavior)
• [FEATURE]: Cooldown-aware runtime failover on rate limits and temporary provider failures #16867: Cooldown-aware runtime failover on rate limits and temporary provider failures (related: retry on rate limits)
• Provider timeout value is not respected #15582: Provider timeout value is not respected (related: provider timeout handling)

If you believe this was flagged incorrectly, please let a maintainer know.

[github-actions]

This issue has been automatically closed because it was not updated to meet our contributing guidelines within the 2-hour window.

Feel free to open a new issue that follows our issue templates.