Broadcom pi zero2w neopixel misbehaving/crash fix

[RetiredWizard]

These changes fix #5947. I've tested on both the zero2w and the pi4b (The 4b didn't exhibit the original issue) and the boards now behave properly with 1 to 30 pixels. I've run a 30 pixel animation on the zero2w for 8+ hours without a hang.

Original corrupted rgb transmission:

Early Fix, not using transmission queue:

This PR:

[RetiredWizard]

I just tested this on a zero w rather than the zero2w and it works but it looks like the code is getting stuck more frequently than the other tested board in the loop that checks for an empty queue. I'm not sure that's a problem if I reduce the loop abort counter but I want to test on the three models I have. I'll update once I've finished testing.

[RetiredWizard]

Tested on 4b, zero2w and zero w. Runs without glitches or crashes.

[tannewt]

Please use && in your conditions instead of &. & is a bitwise test and && is boolean logic.

[RetiredWizard]

So I misunderstood what COMPLETE_MEMORY_READS was doing and clearly over used it in the update for the zero w. I've pulled all the calls I added for that update out and I'm testing it now. The only one remaining that I added was at the very end of the code just before the PWM function of the pin is turned off. I feel like that one might be needed but I guess I'll do some tests.

edit: It doesn't look like that last memory barrier was needed either....

[tannewt]

Thank you!