Add the two missing ASK line codes so an unrecognised tag yields its raw bits
instead of reading as "nothing":
- ask.nrz_bits: NRZ/direct demod. Keys the bit clock off the shortest run (not
2x a half-bit as in Manchester), resamples, yields both polarities/phases.
- ask.biphase_bits: biphase/differential-Manchester, reusing dsp.biphase_raw_decode
(the FDX-B path).
- capture.demod_raw: runs both, finds the shortest clean repeating frame, returns
{period, bits, hex} per encoding + the recovered clock. Sits BELOW the credential
decoders — it names no format, just hands back the bits.
- identify_lf falls back to demod_raw when no credential decodes off the air, so a
strong-but-unknown LF signal surfaces its frame instead of None.
Hardware-hardened: on a live non-NRZ/biphase tag the first cut reported a spurious
16-bit "frame" — biphase-decoding a non-biphase signal emits a short prefix then a
constant run that trivially repeats at any period. _frame_period now rejects
near-constant streams and finds the smallest *true* period (so 0101 fills read as
period 2 and are rejected, not reported at min_p). Regression tests cover all three.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
a990d725ef