fix(semantic): enable CUDA on WSL and report backend

[luca-ctx]

Summary

• recognize WSL2 GPU forwarding as NVIDIA acceleration evidence
• expose the active semantic backend and compute mode in ctx status
• preserve native Linux detection and avoid daemon pings when semantic runtime is inactive

Validation

• scripts/bazel-affected.sh origin/main (86/86)
• focused semantic-model, daemon-cli, and rustfmt checks

Fixes #717