fix pio_encode_mov in release mode

[kilograham]

fixes #3068

[cadwal]

I can confirm this fixes the issue.

On a Pimoroni Presto (RP2350B, ST7701 panel driven by PIO + DMA) the display comes up lit and permanently black on 2.3.0, and renders normally on 2.1.1.

The driver clears the OSR before restarting a frame:

pio_sm_set_enabled(pio, sm, false);
pio_sm_clear_fifos(pio, sm);
pio_sm_exec_wait_blocking(pio, sm, pio_encode_mov(pio_osr, pio_null));
pio_sm_exec_wait_blocking(pio, sm, pio_encode_out(pio_null, 32));

In Release on 2.3.0 the first instruction encodes as MOV PINDIRS, NULL, so the OSR is never cleared. With autopull enabled the following OUT NULL, 32 stalls on a TX FIFO that was just cleared with its DMA aborted, and pio_sm_exec_wait_blocking never returns.

One small thing in the new test — the pio_pindirs_mov check is guarded by #if PICO_PIO_VERSION > 1, but pio_pindirs_mov is defined under #if PICO_PIO_VERSION > 0. > On RP2350 (PIO version 1) that assertion compiles out, so the case closest to the original #2946 fix is not actually covered. Looks like it should be > 0?

[will-v-pi]

LGTM