improve SAMx5x NeoPixel timings

[dhalbert]

Fixes #4102.

This seems to prevent incorrect colors on the first NeoPixel in a string on SAMx5x processors. This is especially noticeable on 5v-level-shifted pins, such as D5 on ItsBitsy M4.

Timings were revamped by empirically changing the delay counts and observing the waveforms. The current timings in ns are below, and can vary by a couple of ns:

zero: 402 high / 834 low
one: 836 high / 402 low

This is close to the RP2040 timings, which uses PIO for precise timings
input zero: 416 high / 832 low
input one: 832 low / 416 high

Note that the far-end Neopixel, which regenerates the timings, is not like this. These far-end timings was observed when driving with an RP2040
output zero: 268 high / 978 low (varies slightly by a few ns)
output one: 698 high / 550 low (varies slightly)

[dhalbert]

Setting to draft while awaiting forum user tests, so it doesn't get merged before that testing.

[kattni]

Tested with some odd strips on ItsyBitsy M4 on pin D5. One strip had the first pixel issue when lighting up red. The other two had no issues. Approving with the caveat that the weirder strips will still need to be tested and the code perhaps refined.

[dhalbert]

closed in favor of #6312