feat(gpu): introduce GPU request spec

[elezar]

Summary

Replace the sandbox GPU boolean/string fields with an optional GPU request spec and unify CDI GPU device ID handling across Docker and Podman.

Related Issue

None

Changes

• Centralized per-driver GPU request validation before changing the spec shape.
• Simplified VM GPU request derivation so one optional value drives GPU assignment.
• Added optional GpuRequestSpec with repeated device_id fields in public and driver protos.
• Shared CDI GPU device ID mapping through openshell-core and wired Docker/Podman through it.
• Updated Kubernetes and VM GPU request validation, CLI/server mapping, tests, and architecture docs.
• Added CDI GPU e2e coverage for default GPU requests and specific device requests via nvidia.com/gpu=0, with Docker and Podman GPU task wiring.

Testing

• mise run pre-commit passes
• cargo test --manifest-path e2e/rust/Cargo.toml --features e2e --test cdi_gpu --no-run
• cargo clippy --manifest-path e2e/rust/Cargo.toml --all-targets -- -D warnings
• Unit tests added/updated
• E2E tests added/updated (if applicable)

Checklist

• Follows Conventional Commits
• Commits are signed off (DCO)
• Architecture docs updated (if applicable)

[copy-pr-bot]

Auto-sync is disabled for draft pull requests in this repository. Workflows must be run manually.

Contributors can view more details about this message here.