Add standalone two-sensor Pulse Transit Time web lab

[yury-g]

What changed

This PR now keeps the two products explicitly separate:

• Signal Coach remains a one-sensor A0 coach with its own page, serial adapter, sender, documentation, and validation lane.
• Pulse Transit Time Lab is a standalone two-sensor A0/A1 research app under docs/pulse-transit-time/, with its own page, strict PTT1 protocol, 500 Hz / 250000-baud sender, tests, documentation, and physical gate.
• PTT deliberately imports only the transport-neutral signal-coach-core.mjs detection algorithm. No Signal Coach UI, protocol adapter, firmware identity, or validation result is shared.

The standalone PTT app adds:

• synchronized A0 proximal / A1 distal acquisition;
• independent channel quality, threshold, clipping, warmup, and pairing checks;
• one-to-one same-cycle pairing in a configurable 5–300 ms window;
• live threshold/window tuning that resets analysis without dropping the serial connection;
• strict canonical unsigned-decimal PTT1,timestamp_us,proximal,distal parsing;
• targeted, idempotent teardown for manual disconnect, EOF/read failure, and the selected-port disconnect event;
• visible simulated-bench labeling that is absent from the normal URL.

The PR also retains the library-side fixed/adaptive-threshold support and stale-cycle PTT pairing hardening, and prevents the one-sensor Signal Coach sender from emitting a startup catch-up burst after its setup delay.

Verification

• 6/6 Node protocol, core, and source-pack suites passed.
• Headless browser lifecycle checks passed for EOF, read failure, manual disconnect, selected-device disconnect, unrelated-device disconnect, and live PTT tuning.
• Arduino UNO R4 WiFi compiles passed:
  • Signal Coach sender: 52,048 bytes flash / 6,744 bytes RAM
  • standalone PTT sender: 52,288 bytes flash / 6,744 bytes RAM
  • library PTT example: 56,568 bytes flash / 6,848 bytes RAM
• Checked-in and tutorial Signal Coach sender bodies match, including startup scheduling.
• The PTT page's embedded sender matches the checked-in sketch.
• Local links and git diff --check passed.
• Independent adversarial review found no remaining P1–P3 issues.

Physical HOLD gate

This remains a draft. Browser simulation, compilation, and transport behavior are not physiological PTT evidence.

Before merge, publication, issue #202 closure, or customer notification, record a real two-sensor run showing:

• A0 proximal and A1 distal with shared 5V/GND;
• both channel quality states sustained at GOOD without clipping;
• repeatedly increasing same-cycle accepted-pair count;
• a reasonably stable median over a sustained placement;
• a visibly fresh result after disconnect/reconnect;
• exact placement, attachment, threshold, window, duration, and raw result notes.

PTT here is an educational timing experiment. It does not estimate blood pressure and is not for diagnosis or treatment.