Address GitHub Copilot review comments on VM testing PR

[Copilot]

Addresses all 10 automated review comments on the VM integration testing infrastructure PR.

Critical Fixes

• Remove unused microvm dependency - Dropped microvm flake input; tests use pkgs.testers.runNixOSTest, not microvm.nix
• Fix TUN interface detection for Linux - Parse actual interface name from logs instead of hardcoded macOS utun*:

  tun_match = re.search(r"TUN device created: (\S+)", log_output)
  tun_name = tun_match.group(1)
  machine.succeed(f"ip link show {tun_name}")

• Fix concurrent receiver output - Separate stderr to prevent hash file corruption:

  f"> /tmp/hash_{port}.txt 2> /tmp/recv_{port}.log &"

Documentation & Quality

• Update docs to match implementation - doc/vm-testing.md and AGENTS.md now describe NixOS test framework
• Optimize CI workflow - Remove redundant nix flake check and duplicate builds; separate VM/non-VM checks
• Fix module test key generation - Check if key exists before generating; restart service after key creation to avoid race
• Python socket handling - Use context managers in receive_tcp.py for proper resource cleanup

Files Changed

• .github/workflows/test.yml - CI optimization
• AGENTS.md, doc/vm-testing.md - Documentation corrections
• flake.nix, flake.lock - Remove microvm dependency
• tests/vm/helpers/*.py - TUN detection, key generation, socket handling
• tests/vm/*-test.nix - TUN detection, output redirection

---

✨ Let Copilot coding agent set things up for you — coding agent works faster and does higher quality work when set up for your repo.