Harden hardware-free validation, packaging, and demo evidence - #10

Summary

• reject malformed measurement/reference arrays instead of silently sorting, deduplicating, or filling them
• verify the deterministic fixture -> reference correction -> MAT/CSV/TXT export -> reload workflow end to end
• improve fixture-mode Bode plotting, neutral AWG/OSC state, reference receipts, and scroll-safe operator layout
• run the test suite across Linux, macOS, and Windows, then build and smoke-test a Windows PyInstaller one-folder package
• refresh the real Tk point-replay capture and public validation/evidence documentation

Validation

• local Python 3.11 and 3.13 suites: 82 tests each
• local Ruff, compile, Markdown link, diff, source package-smoke, and 112-frame real-Tk capture integrity checks
• clean local PyInstaller one-folder build plus frozen-binary package smoke
• GitHub Actions run 29435713491:
  • Ubuntu Python 3.10 and 3.13: passed
  • macOS Python 3.11: passed
  • Windows Python 3.11: passed
  • Ruff: passed
  • Windows one-folder build and packaged smoke: passed
• downloaded Windows receipt confirms source=mock_fixture, point_count=72, three exported artifacts, and live_hardware_used=false

Validation boundary

This PR contains hardware-free software, persistence, package, and real-UI evidence. The fixture replay and package smoke never construct production AWG/oscilloscope adapters. It does not claim live VISA discovery, connected-instrument behavior, metrological accuracy, hardware shutdown timing, or production safety certification.