Add Requesty adapter (OpenAI-compatible)

[Thibaultjaigu]

This adds a dedicated requesty adapter, mirroring the existing OpenRouter adapter as closely as possible.

Requesty is an OpenAI-compatible gateway (base URL https://router.requesty.ai/v1) that uses provider/model naming — the same convention as OpenRouter — so it reuses the existing client.OpenAICompatClient with no new client code.

Changes (mirror of the OpenRouter wiring)

• New pkg/adapters/requesty.go — RequestyAgent mirroring OpenRouterAgent: reads REQUESTY_API_KEY, targets https://router.requesty.ai/v1, registers via agent.RegisterFactory("requesty", NewRequestyAgent) in init(). Same Initialize/HealthCheck/SendMessage/StreamMessage/buildConversationHistory logic.
• New pkg/adapters/requesty_test.go — unit + integration tests mirroring openrouter_test.go (integration tests skip unless REQUESTY_API_KEY is set).
• cmd/model_validation.go — adds the requesty entry (Supported + Required), matching openrouter.
• internal/providers/providers.json — minimal requesty provider entry (id, OpenAI type, endpoint, a few known models) so cost/validation lookups resolve; unknown models still fall through to the existing "not in registry → warning" path.
• examples/requesty-conversation.yaml, examples/requesty-solo.yaml — mirror the OpenRouter examples.
• README.md — adds Requesty to the supported-agents list and the model-support table.

Testing

• go build ./..., go vet ./..., and gofmt -l are clean.
• go test -short ./pkg/adapters/... ./internal/providers/... passes (the embedded providers.json loads correctly).
• Live integration tests against https://router.requesty.ai/v1 pass: HealthCheck and SendMessage with openai/gpt-4o-mini returned a real completion ("Test successful.").

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