Feature request: Support hosted/cloud embedding providers (OpenAI, OpenRouter, etc.)

[daronyondem]

Hey! Love that qmd is all-local, but it would be great to have an optional cloud embedding backend for cases where:

• The local machine lacks GPU (CPU-only embedding is slow)
• Higher quality embeddings are desired (e.g. OpenAI text-embedding-3-small/large)
• Users already have API keys for OpenRouter, OpenAI, etc.

Proposed approach:

• Add an optional config (env var or config file) like QMD_EMBED_PROVIDER=openai and QMD_EMBED_API_KEY=sk-...
• Default remains local GGUF models (no breaking change)
• When configured, qmd embed calls the hosted API instead of running the local model
• Providers to consider: OpenAI, OpenRouter, Cohere, Voyage AI

This would keep qmd's simplicity while unlocking better embeddings for users who want them. The local-first default stays intact.

Thanks for building this!