Add Requesty as an LLM provider

[Thibaultjaigu]

Adds Requesty as a first-class LLM provider, mirroring the existing OpenRouter provider end-to-end.

Requesty (https://requesty.ai) is an OpenAI-compatible LLM router — base URL https://router.requesty.ai/v1, GET /v1/models in OpenAI shape, Authorization: Bearer auth, optional HTTP-Referer/X-Title headers, provider/model naming.

Changes:

• src/server/infra/http/requesty-api-client.ts (new): mirror of openrouter-api-client.ts (base URL + names differ; same headers/fetch/cache/normalize logic).
• src/server/infra/http/provider-model-fetchers.ts: RequestyModelFetcher mirroring OpenRouterModelFetcher.
• src/server/infra/http/provider-model-fetcher-registry.ts: case 'requesty' + import.
• src/server/infra/provider/provider-config-resolver.ts: case 'requesty' (routed through the generic OpenAI-compatible providerApiKey/providerBaseUrl path; OpenRouter's dedicated openRouterApiKey field is a legacy special case).
• src/server/core/domain/entities/provider-registry.ts: requesty catalog entry (label, baseUrl, defaultModel openai/gpt-4o-mini (verified live), envVar REQUESTY_API_KEY, headers, requiresApiKey) — drives env auto-detection.
• src/agent/infra/llm/providers/requesty.ts (new) + index.ts registration + src/agent/core/domain/llm/registry.ts (requesty in the openai providerType list).

tsc --noEmit passes clean (verified after npm ci).

I work at Requesty. This mirrors the existing OpenRouter provider as closely as possible. Happy to adjust or close it if it's not a fit.