feat(providers): validate model connectivity before save

[slicenferqin]

Summary

This PR adds a save-time connectivity validation path for provider setup/edit flows, so users can catch provider/model misconfiguration before returning to chat.

• Extend provider:validateKey to accept optional model and validate model connectivity for OpenAI-compatible providers.
• Reuse stored API key when editing provider base URL/model without entering a new key.
• In Setup and Provider Settings, pass model into validation calls so we test both key + model before save.
• Add explicit Test Connection button and gate saving on successful connectivity test:
  • Setup: users must run test and pass before save is enabled.
  • Settings/Add Provider: same gating behavior for consistency.
• For OpenRouter, run auth check first, then model connectivity check (when model provided).

Why

Users can currently save provider config successfully, but only discover model/baseURL issues after sending a chat message. This adds an earlier fail-fast check in config flows.

Validation

• pnpm -s typecheck
• JSON parse check for updated locale files

[slicenferqin]

已同步最新 main 并解决冲突，相关更新已经推送到当前 PR 分支。\n\n本次合并主要处理了 src/components/settings/ProvidersSettings.tsx 和 src/pages/Setup/index.tsx 的冲突，并保留了这条 PR 原本的“保存前做连接测试 / 模型连通性校验”逻辑，同时兼容了 main 上后续引入的 provider UI/鉴权模式调整。\n\n本地验证已通过：\n- pnpm run lint\n- pnpm run typecheck\n- pnpm run test\n\nIf helpful, I can also split or further simplify any part of this PR for easier review.\n\nI’ve merged the latest main into this branch and resolved the conflicts.\n\nThe conflict resolution mainly covered src/components/settings/ProvidersSettings.tsx and src/pages/Setup/index.tsx. The PR’s original behavior is preserved: provider/model connectivity must be validated before save, while also staying compatible with the newer provider UI and auth-mode changes from main.\n\nLocal validation passed:\n- pnpm run lint\n- pnpm run typecheck\n- pnpm run test\nEOF