Add LLM fallback clients, daily quota tracker and quota-board UI

[Lzxpan]

Motivation

• Support multiple LLM backends (local vLLM, Mistral, Groq) and automatically failover between configured API keys when a provider is rate/quotas limited.
• Provide per-model daily usage accounting so users can see and persist daily quotas and errors to avoid unexpected rate limits.
• Expose new model/API key settings and a simple UI to view today’s engine usage without modifying existing workflows.

Description

• Added a daily usage tracking subsystem in DailyQuotaTracker.cs that records per-provider/model request counts, token usage, last error, and default daily/RPM limits and persists to usage_daily.json via DailyQuotaTracker.LoadOrCreate and Save.
• Implemented OpenAI-compatible client and multi-provider credential model in LlmClients.cs with OpenAiCompatibleClient, LlmCredential, LlmProvider, and error policy helpers; added TranslateTextWithFallbackAsync and TranslateImageWithFallbackAsync in Form1.cs to try configured credentials sequentially and record results to the quota tracker.
• Extended UI and settings: added input fields and settings entries for ApiKey_Pixtral, ModelName_Pixtral, ApiKey_MistralPixtral, ModelName_MistralPixtral, ApiKey_Llama4, and ModelName_Llama4 in Form1.Designer.cs, Properties/Settings.*, and App.config defaults, and wired them into Form1 load/save logic.
• Added a lightweight QuotaBoardForm inside Form1.cs and a btnQuotaBoard button to show the current day’s usage snapshot returned by DailyQuotaTracker.GetSnapshot, and added logic to persist usage on exit and on each recorded request.
• Minor UX change: Form1_FormClosing now prompts the user with choices to exit or minimize to tray before saving settings and usage.

Testing

• No automated tests were added or executed as part of this change (no CI test suite changes).

---

Codex Task

[chatgpt-codex-connector]

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: f567fd966e

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

• Open a pull request for review
• Mark a draft as ready
• Comment "@codex review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".

[chatgpt-codex-connector]

Skip implicit local vLLM credential when not configured

This block always adds the local Pixtral credential whenever a model is selected, and Form1_Load always selects one by default, so users who only configured Mistral/Groq still hit http://127.0.0.1:8000/v1 first. If local vLLM is not running, the first request fails with a non-quota error and TranslateTextWithFallbackAsync/TranslateImageWithFallbackAsync return immediately instead of reaching later configured providers, causing translation to fail despite valid remote API keys.

Useful? React with 👍 / 👎.

[chatgpt-codex-connector]

Persist quota data in a user-writable directory

The quota tracker file is created under Application.StartupPath, which is commonly read-only for normal users when the app is installed under Program Files. Because DailyQuotaTracker.Save() suppresses IO exceptions, quota history silently fails to persist in that environment. Using a per-user data path (for example %LocalAppData%) would avoid silent data loss.

Useful? React with 👍 / 👎.