Improve external clock/reset sync (interrupt-based edge capture, no blocking stalls)

[othmar52]

When clocking the Muxlicer from an external CV clock + reset, the playhead
was often slightly late and occasionally ended up a full step behind until
the next reset. This PR removes the causes:

• External clock via pin change interrupt (PCINT20). The clock input
  (PD4) was polled once per loop() pass, so edge detection jittered by
  the loop period (a few hundred microseconds, dominated by blocking
  analogRead calls) and pulses shorter than a blocking call were missed
  entirely. The ISR only stores a micros() timestamp and a flag; all
  processing stays in read_clock(). Timing references (clock out, gate
  windows, step advance) now use the captured edge timestamp, so the
  measured clock period is free of polling jitter.

• Reset via pin change interrupt (PCINT11). Same treatment for the
  reset input (PC3/A3); short reset pulses can no longer be missed. The
  processing order within the loop (reset before clock) is unchanged.

• EEPROM writes deferred while running. EEPROM.write blocks ~3.3 ms
  per byte; persisting tempo + clock-in mult (up to 8 bytes) froze the
  loop for up to ~26 ms. These writes were triggered 3 s after the last
  encoder move, on tap-button release and when the external clock was
  unplugged — i.e. typically while playing, stalling steps/gates/clock out
  audibly and (before the interrupt capture) even losing pulses. All
  writes now go through service_pending_EEPROM_writes(), which performs
  them only while the sequencer is stopped; pending writes stay flagged
  until then. Note: settings changed while running are only persisted
  once the sequencer is stopped before power-down.

• Address CV throttled to 500 Hz. analogRead blocks ~112 µs and ran
  every loop pass; reading the address CV every 2 ms is still far faster
  than any musically relevant CV change and roughly halves the loop
  period.

No hardware changes required (ATmega328P pin change interrupts).

Tested on a Muxlicer (PCB v1.3) clocked from an Arturia BeatStep Pro
(clock + reset): with clock-in mult/div at 0, ~50 stop/start cycles were
perfectly in sync (previously the playhead lagged audibly and
intermittently lost steps).

Commits

1. Capture external clock via pin change interrupt
2. Defer EEPROM writes while the sequencer is running
3. Throttle address CV reads to every 2 ms
4. Capture reset input via pin change interrupt