feat: add Requesty as an OpenAI-compatible LLM provider

[Thibaultjaigu]

Summary

Adds Requesty as a first-class LLM provider. Requesty is an OpenAI-compatible
LLM gateway, so this provider is built on the existing OpenAI-compatible model path —
the same way the local and openrouter providers are constructed — rather than
introducing a new dependency or a provider-specific model class.

Models follow the provider/model naming convention (e.g. openai/gpt-4o-mini),
mirroring OpenRouter. The provider resolves to litellm's OpenAI-compatible path via the
existing {provider}/{model_name} identifier plus a base URL, so no new dependency is
required (the OpenAI-compatible client already in the tree is reused).

Configuration

• REQUESTY_API_KEY — API key sent as a Bearer token
• REQUESTY_MODEL — default openai/gpt-4o-mini (uses provider/model naming)
• REQUESTY_BASE_URL — default https://router.requesty.ai/v1

Wiring sites updated

• deadend_cli/deadend_agent/src/deadend_agent/config/settings.py — new requesty_api_key,
  requesty_model (default openai/gpt-4o-mini), and requesty_base_url
  (default https://router.requesty.ai/v1) settings fields, mirroring the OpenRouter/Local
  fields; provider docstring updated to document Requesty.
• deadend_cli/deadend_agent/src/deadend_agent/models/registry.py — provider docstring updated
  to include Requesty.
• deadend_cli/src/deadend_cli/component_manager.py — added requesty to the provider
  help/enable lists (openai, anthropic, gemini, openrouter, requesty, local).
• deadend_cli/.env.example — added commented REQUESTY_API_KEY / REQUESTY_MODEL /
  REQUESTY_BASE_URL example block.
• README.md — added Requesty alongside OpenRouter in the supported-providers list.
• cli/deadend/src/app/deadend-app.ts — added requesty to the setup-wizard provider example hint.

Verification

• ast.parse syntax check passed on every edited Python file (settings.py, registry.py,
  component_manager.py).
• Loaded settings.py in isolation and confirmed the new fields resolve with the expected
  defaults: requesty_model='openai/gpt-4o-mini', requesty_base_url='https://router.requesty.ai/v1',
  requesty_api_key=None when unset.
• Live chat-completion smoke test against https://router.requesty.ai/v1/chat/completions with
  model openai/gpt-4o-mini returned HTTP 200 and a valid completion (model: gpt-4o-mini-2024-07-18),
  confirming the base URL, auth header, and provider/model naming work end-to-end.
• I could not run the full uv sync test suite in this environment: the workspace declares a git
  submodule (simple-python-interpreter-sandbox) that is not checked out here, so uv sync fails
  on a missing workspace member's pyproject.toml — unrelated to this change. The changes are
  documentation/config/string additions plus three new settings fields, with no behavioral changes
  to existing providers.

Links

• https://requesty.ai
• https://docs.requesty.ai

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.