refactor: update conversation handling and improve message state management

[chenyme]

Summary

支持模型生成期间的分支切换、历史消息重试和用户消息编辑。

本次将聊天运行时从单一 pending exchange 重构为按 runID 隔离的并发状态，确保每个生成任务拥有独立的流缓冲、推理内容、定时器和取消控制器。同时修复并发完成顺序、队列父节点、历史分页、服务端 pending 状态和 OpenAI Responses 会话状态覆盖问题。

Change type

• Bug fix
• Feature
• Documentation
• Refactor
• Configuration / deployment
• Security hardening
• Other

Affected areas

• Frontend / UI
• Backend / API
• Authentication / authorization
• Conversations / streaming
• Files / RAG / extraction
• Model routing / providers
• MCP / tools
• Billing / payments
• Admin console
• Deployment / Docker / configuration
• Documentation

Verification

• cd frontend && pnpm exec tsc --noEmit
• cd frontend && pnpm lint
• cd backend && go test ./...
• git diff --check
• Frontend build not run by design.

Screenshots, API examples, or logs

Not applicable. This change does not introduce a new standalone UI surface.

Configuration, migration, and compatibility notes

• No API contract, database schema, configuration, or migration changes.
• Normal composer submissions continue to queue while generation is active.
• Historical retry/edit branches may run concurrently with active generation.
• Concurrent generation is limited to three unique active runs on the current page.
• Non-default branches do not overwrite the conversation’s main OpenAI Responses state.
• After page reload, multiple pending runs are recovered sequentially.

Documentation

• Documentation is not needed for this change.
• Documentation was updated.
• Documentation still needs to be updated.

Security and privacy

• No secrets, tokens, credentials, local config, or personal data are included.
• User data access remains scoped by authenticated user context unless an admin-only path explicitly requires broader access.
• No authentication, authorization, billing, or permission behavior was changed.

Checklist

• I searched existing issues and pull requests.
• Changes are focused and do not include unrelated refactors.
• Tests or static verification were run where practical.
• User-facing behavior and compatibility impact are documented.
• Generated artifacts are included only when this project explicitly requires them.
• Caches, build output, .pyc files, .env files, and local storage data are not committed.