feat: add optional Headroom token saver

[CarmeloCampos]

Summary

• Adds optional Headroom compression via an external /v1/compress proxy before provider dispatch
• Adds dashboard and CLI toggles plus configurable Headroom proxy URL
• Documents local and Docker setup while keeping Headroom unbundled

Why

Closes #1871.

This lets users run:

Client -> 9Router -> Headroom /v1/compress -> 9Router -> provider

9Router keeps routing, auth, fallback, usage logging, and provider translation. Headroom only handles context compression.

Benefits

• Optional extra token savings on top of RTK
• No new runtime dependency or bundled Python process
• Docker-friendly via configurable URL, e.g. http://headroom:8787 or http://host.docker.internal:8787
• Fail-open behavior: if Headroom is down or errors, the original request is sent unchanged

Testing

• NODE_PATH=tests/node_modules tests/node_modules/.bin/vitest run --config tests/vitest.config.js tests/unit/headroom.test.js --reporter=verbose -> 6 passed
• npm run build -> TypeScript completed

[GabrielCastroV]

Looks pretty good actually 👀

[decolua]

Thanks a lot for this contribution, @CarmeloCampos! 🙏

The optional Headroom token saver integration is a great addition. We've merged it locally and built on top of it with:

• Managed start/stop of the local Headroom proxy directly from the dashboard
• CLI install + Python (>=3.10) detection and proxy /health probing
• One-click Start/Stop UI in Token Saver, no manual config needed
• Claude <-> OpenAI shape conversion for /v1/compress via the 9Router translators

Closing this PR as the work has been integrated. Really appreciate the original implementation that made it easy to extend. 🚀