DigitalInput::wait() optimizations

[necroware]

Trying to get some higher performance by reducing the time spent in the DigitalInput::wait() function. Hopefully this will provide a little bit more time to read the bits from the data lines in time. Unfortunately even higher optimization would require a complete rewrite of the data reading structures. This is a try to fix the bug but keep the code simple.

[creopard]

Unfortunately this fix in DigitalPin.h does not seem to work.
The lines:

if (edge == Edge::falling) {
  return waitImpl(timeout, [](uint8_t a, uint8_t) {return a;});
}
if (edge == Edge::rising) {
  return waitImpl(timeout, [](uint8_t, uint8_t b) {return b;});
}
// edge == Edge::rising
return waitImpl(timeout, [](uint8_t a, uint8_t b) {return a|b;});

will break the Sidewinder 3D implementation and the Sidewinder will then be recognized as 2 axis, 2 button joystick

replacing that section
uint8_t read() const {
with
bool read() const {

and replacing the coding with the original implementation (and the renamed method "read()") fixes that again:

auto last = read();
for (; timeout; timeout--) {
  const auto next = read();
  if (last == next) {
    continue;
  }
  switch (edge) {
    case Edge::falling:
      if (last > next) {
        return timeout;
      }
      break;
    case Edge::rising:
      if (last < next) {
        return timeout;
      }
      break;
    case Edge::any:
      return timeout;
  }
  last = next;
}
return 0u;

This may also be applicable for the Logitech implementation.
Sorry, I had no time to investigate further, but at the current state the Sidewinder 3D implementation is broken (the last release v0.2.1 is of course not affected by this)