feat: preserve historical images in multi turn conversations

[chenyme]

Summary

支持在多轮对话中按原始用户消息传递历史图片，使模型能够继续分析、比较或追问此前上传的图片。

• 将历史用户图片恢复到其原始消息位置，避免图片与消息上下文错位。
• 保留最近 10 张历史图片，并预留相应的上下文 Token 预算。
• 历史图片上下文总大小限制为 20 MiB。
• 图片缺失、读取失败或超过限制时返回明确错误，不再静默忽略。
• 修复图片缩放后实际编码与 MIME 类型不一致的问题。
• 增加有界 LRU 缓存，减少历史图片的重复读取和缩放。
• 保持 OpenAI Responses previous_response_id 的上下文指纹稳定。
• 不预判模型的视觉能力；不支持图片时保留上游的正常错误行为。

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 backend && env GOCACHE=/Users/project/DEEIX-Chat/.gocache go test ./...
• git diff --check

Screenshots, API examples, or logs

Not applicable. This change affects backend conversation context construction.

Configuration, migration, and compatibility notes

• No database migration or public API change is required.
• Historical image context retains the most recent 10 images.
• Prepared historical images are limited to 20 MiB per request.
• The in-process prepared-image cache is bounded to 64 entries and 64 MiB.
• Models without image support will continue to return their normal upstream error.

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.
• Security-sensitive behavior was reviewed, including file ownership, object storage access, and image context limits.

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, deployment steps, API contracts, or configuration changes 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.