docs: add voice input via MCP server guide - #42769

Thanks for updating your PR! It now meets our contributing guidelines. 👍

First-time contributor here — the CI workflows are pending maintainer approval. Could a maintainer approve the run and take a look? Happy to adjust placement (docs vs in-repo packages/) if preferred. Thanks

AI code review — automated review; reference only.

Scope: docs-only — new packages/docs/essentials/voice-input.mdx plus a "Guides" group in packages/docs/docs.json, documenting a third-party local voice-input MCP server.

• Content is clear and technically coherent: MCP config snippet matches opencode's local-server mcp config shape; prerequisites (sox/ffmpeg + faster-whisper) and env-var table are consistent.
• Policy question for maintainers more than a correctness one: this grants official-docs placement (new top-level "Guides" group) to a personal npm package (@yuvarjbhado/voice-mcp). Existing third-party integrations live in the ecosystem listing instead; consider whether that's the right home, or at least whether the "Guides" group should be reserved for first-party content.
• Note the existing essentials/*.mdx pages (code, images, markdown…) are not referenced from the current docs.json navigation, so this PR's nav entry will be the only essentials page wired up — verify the built site renders it where intended.
• Minor: "Configure" and "Configuration" as consecutive section titles is redundant; rename one (e.g. "Environment variables").
• Privacy claims ("nothing sent to a remote server") are the package author's own and unverifiable from this repo — fine for docs, but maintainers are implicitly endorsing them by merging.