Pico programmable input/output (PIO) capture/backpressure — complete experiment + host checks

Physical fixture proposed, not built: original RP2040 Pico, official RPI_PICO
MicroPython 1.26.0, reported system clock 125000000 Hz. Universal Serial Bus (USB) power only.
GP16 physical 21 -> 1 kilohm series -> GP17 physical 22 (input, internal pull-down).
Do not attach an external generator or second output. Power off to move wires.
See pio-functional.svg/png for the complete functional diagram.
Save pico_pio_backpressure.py on Pico. Execute explicitly once:
  scope = {}
  exec(open('pico_pio_backpressure.py').read(), scope)
  scope['main'](blocking=True, hold_ms=20)
Then compare blocking=False, and hold_ms=0 controls. No autostart supplied.

Host checks (not target firmware build/execution):
  python3 -m venv .venv
  .venv/bin/python -m pip install -r requirements-host.txt
  .venv/bin/python test_capture.py
Optional constructed trace:
  .venv/bin/python host_support.py
The pinned MicroPython assembler source and host stub constants encode the actual
six-instruction program. The third-party opcode emulator is 0.88.0. Its ideal
input bit 0 represents in_base GP17; it does not validate pin configuration, synchronization,
physical timing or the FDEBUG register. RXSTALL condition comes from the data-
sheet rule applied to full PUSH attempts. No measured samples are in this archive.
Four queued first-in, first-out (FIFO) words are distinct from a pending fifth input shift register (ISR) word. Contiguous tokens
never prove complete physical event history. Do not invent event timestamps.
observations-blank.csv is header-only; retain constructed results separately.
