Add Alpamayo ModelOpt FP8 and Edge export guides

[micwill755]

Summary

• add a standalone Alpamayo 1.5 ModelOpt FP8 guide that follows the official NVlabs recipe exactly: pinned uv environment, full-model PhysicalAI calibration, default mtq.compress checkpoint save, ModelOpt state verification, and minADE evaluation
• add the stacked quantize-to-Edge-export guide using the same compressed alpamayo1.5_fp8_calib100 checkpoint
• document the separate export environment, Edge-LLM plugin build, vision/language/action export, parity validation, artifacts, checklist, and troubleshooting
• cross-link both guides from the general quantization and Edge exporter documentation

Validation

• documentation pre-commit and typo checks pass
• Sphinx HTML build succeeds with gallery execution disabled
• both Alpamayo pages render without warnings

Stack

This PR is stacked on #4732 (alpamayo-edge-exporter-support).

Related pull requests

• Added alpamayo fp8-woq torch ao example #4731 — TorchAO FP8 weight-only Alpamayo expert workflow
• Alpamayo Edge exporter support #4732 — Alpamayo unified Edge exporter support