feat: add configurable provider RPM limiting

[BelugaRex]

Summary

• add global default and per-provider RPM configuration; set rpm: 0 to disable a provider's limiter
• use a token bucket at the logical chat-request entry point so HTTP, SSE, and Responses WebSocket traffic share the same limit
• serialize concurrent token acquisition FIFO to prevent negative balances or duplicate consumption
• display the post-acquire token snapshot with two decimal places and document the per-window scope

Validation

• npm run compile
• npm run l10n:check
• concurrent RateLimiter(60) harness: 52 requests completed FIFO; capacity was 48; the final four completed at approximately 1-second intervals; no snapshot became negative

Fixes #242

[chatgpt-codex-connector]

You have reached your Codex usage limits for code reviews. You can see your limits in the Codex usage dashboard.

[BelugaRex]

希望大佬见谅，我是vide的，但是我有本地测试过，好像是能有效果的。（比较悲哀的是，就算我卡rpm在4，也会被老黄429，唉）

[Copilot]

Pull request overview

This PR adds configurable RPM (requests per minute) rate limiting for logical chat requests, with a global default and per-provider overrides, and applies the limiter at the service entry point so all transports (HTTP/SSE/WebSocket) share the same limit within a VS Code window.

Changes:

• Introduces a token-bucket RateLimiter with FIFO token acquisition and a createRateLimiter helper.
• Wires rate-limit token acquisition + status logging into UnifyChatService and the various provider clients via ApiProvider optional methods.
• Adds UI/config plumbing (schema, forms, descriptions) and localization strings for global and per-provider rateLimit.rpm.

Reviewed changes

Copilot reviewed 25 out of 26 changed files in this pull request and generated 3 comments.

Show a summary per file

File	Description
src/ui/screens/timeout-form-screen.ts	Adds rate-limit RPM field editing to the timeout/network settings UI flow.
src/ui/screens/provider-form-screen.ts	Passes rateLimit into the timeout form route for provider editing.
src/ui/screens/provider-draft-form-screen.ts	Passes rateLimit into the timeout form route for draft editing.
src/ui/router/types.ts	Extends TimeoutFormRoute to include rateLimit.
src/ui/provider-fields.ts	Includes rate-limit presence/value in provider “timeout” field description string.
src/types.ts	Adds rateLimit?: RateLimitConfig to ProviderConfig with documented semantics.
src/service.ts	Acquires a rate-limit token before transport selection and logs the limiter snapshot.
src/rate-limit.ts	New token-bucket rate limiter implementation + config type + factory.
src/logger.ts	Adds formatting and log suffix for rate-limit token snapshots.
src/config-store.ts	Reads global networkSettings.rateLimit and applies as a provider default.
src/config-ops.ts	Persists rateLimit as part of provider config keys.
src/client/openai/responses-client.ts	Creates and exposes rate limiter for OpenAI Responses provider.
src/client/openai/chat-completion-client.ts	Creates and exposes rate limiter for OpenAI Chat Completions provider.
src/client/ollama/client.ts	Creates and exposes rate limiter for Ollama provider.
src/client/interface.ts	Adds optional getRateLimitStatus / acquireRateLimitToken hooks to ApiProvider.
src/client/google/ai-studio-client.ts	Creates and exposes rate limiter for Google AI Studio provider.
src/client/github-copilot/client.ts	Creates and exposes rate limiter for GitHub Copilot provider.
src/client/anthropic/client.ts	Creates and exposes rate limiter for Anthropic provider.
package.nls.zh-cn.json	Adds Chinese configuration descriptions for rate-limit settings.
package.nls.json	Adds English configuration descriptions for rate-limit settings.
package.json	Adds configuration schema entries for global and per-provider rateLimit.rpm; bumps version.
package-lock.json	Updates lockfile version metadata to match package version bump.
l10n/bundle.l10n.zh-cn.json	Adds UI strings for RPM rate limit editing (Chinese).
l10n/bundle.l10n.json	Adds UI strings for RPM rate limit editing (English).
CHANGELOG.md	Adds v7.12.4 changelog entry covering the rate limiter changes.
.vscode/指北.md	Adds internal guidance doc describing RPM limiter semantics and maintenance boundaries.

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

[BelugaRex]

Addressed Copilot review suggestions

Commit f6982d1 tackles the three Copilot review comments:

1. config-store.ts — Global rateLimit fallback skipped for invalid per-provider config (Medium)

• Added normalizeRateLimitConfig(): invalid/empty per-provider rateLimit (e.g. {} or non-integer rpm) is now normalized to undefined, so it falls back to the global networkSettings.rateLimit default. Explicit rpm: 0 (disable limiter) is still preserved — it does not fall back.

2. rate-limit.ts — getAvailableTokens inflates the logged post-acquire snapshot (Low)

• getAvailableTokens() no longer refills before returning. Since acquire() already refills internally before consuming, the snapshot taken right after an acquisition is the exact post-consumption bucket state instead of being slightly inflated by freshly-earned tokens.

3. service.ts — Rate-limit acquisition is not cancellation-aware (Medium)

• RateLimiter.acquire(signal?: AbortSignal) now accepts an AbortSignal; if aborted before or while waiting, it rejects with an AbortError without consuming a token.
• ApiProvider.acquireRateLimitToken(signal?) and all 6 provider implementations forward the signal.
• service.ts bridges the VS Code CancellationToken into an AbortController passed to acquireRateLimitToken, so a user cancellation while waiting for a token aborts early and doesn't waste a rate-limit slot.

Also updated CHANGELOG.md and .vscode/指北.md, and npm run compile passes clean.