[DOCS] Docs sprint: edge/K3s/ARM/Jetson guides, large-model + quantization guidance, autoscaling tutorial

[Defilan]

What

A batched documentation sprint. These are doc-only deliverables that were sitting in the engineering backlog. Tracking them together so they don't clutter the code milestones.

• Edge deployment guides for K3s, ARM64, and Jetson #8 — Edge deployment guides for K3s, ARM64, and Jetson
• Large model deployment strategies: quantization guidance and single-node multi-GPU optimization #7 — Large model deployment strategies: quantization guidance + single-node multi-GPU optimization
• [FEATURE] Tutorial: metrics-driven autoscaling for InferenceService #478 — Tutorial: metrics-driven autoscaling for InferenceService

Why

No engineering work; pure docs. Grouped so they can be done in one focused pass and tracked separately from the code roadmap. Directly supports the heterogeneous-fleet (edge/ARM/Jetson) and enterprise (autoscaling, large-model) stories.