Add vLLM local OpenAI-compatible server support

[Emilien-Etadam]

Summary

This PR adds support for vLLM as an LLM provider, allowing users to connect
to a local (or remote) vLLM inference server via its OpenAI-compatible API.

Changes

• src/ts/llmProviders/impl/vllmProvider.ts — New provider implementing
  all standard methods (summarize, translate, explain, etc.) plus static
  getModels() to list available models from the server. Supports optional
  API key for secured deployments.
• src/ts/llmProviders/providerFactory.ts — Register vllm in the
  provider map.
• src/ts/helpers/configType.ts — Add vllm: { serviceUrl, model, apiKey }
  to the config interface.
• src/html/partials/options-vllm.html — Options UI with service URL,
  optional API key, and a "Load models" button.
• src/ts/options/optionsVllm.ts — Options handler: loads models on
  provider change or button click.
• src/ts/options/options.ts — Wire up vLLM config save/restore.
• src/html/options.html — Add vLLM to the provider dropdown and include
  the partial.
• src/sass/options.scss — Show the vLLM fieldset when selected; include
  vLLM in the local-LLM timeout warning (consistent with Ollama/LMS).
• src/manifest.json — Add http://localhost/* and http://127.0.0.1/*
  permissions so the extension can reach a local vLLM server.
• src/locales/ — Add EN/IT strings for the vLLM section.

Notes

• vLLM exposes the same /v1/chat/completions and /v1/models endpoints as
  OpenAI, so the implementation closely mirrors the existing LM Studio provider.
• The API key field is optional — most local deployments run without one.
• Default service URL is http://localhost:8000 (vLLM default port).