Select static liburing for Phase 1 native ring

[bright2227]

Summary

• add the Phase 1 native-ring backend decision record
• select the pinned, statically linked liburing route
• record comparable wheel-member size and lifecycle syscall measurements
• document packaging, sanitizer, canonical-module, and raw-spike cleanup follow-up
• link the accepted decision from the roadmap and both spike notes

This is stacked on #15 so the PR contains only the decision work while the static-liburing spike is under review.

Why

Both Phase 1 lifecycle spikes now expose the same boundary. Their lifecycle syscall counts are identical, and WSL2 timing was too noisy to distinguish them. Static liburing increases the compressed wheel-member contribution by about 32 KB in this build, but delegates the correctness-sensitive SQ/CQ shared-memory ordering and kernel-ABI maintenance to the established library.

The decision therefore favors correctness and maintainability while explicitly retaining self-contained source packaging and full-library sanitizer coverage as blockers before production use.

Validation

• .venv/bin/pytest — 33 passed
• .venv/bin/mdformat --check ROADMAP.md docs/phase1-raw-syscall-spike.md docs/phase1-static-liburing-spike.md docs/phase1-native-ring-decision.md
• git diff --check
• uv build --wheel
• uv build --sdist (confirmed the documented missing-vendored-sources gap)
• strace -qq -c over 1,000 construct/close iterations per backend