RP2040: fix off-by-one PWM top issue

[dhalbert]

Fixes #4189.

• The pwmio_pwmout_obj_t top was 32 bits, but it's passed to pico-sdk as 16 bits. Make it 16 bits so the compiler can catch values that are too large. Then needed to force some arithmetic to be 32 bit.

• Maintain top as a 0 to n-1 value. It was getting set to 65536 in one case. Remove self-> top -1 computations.

• Comparison check for low vs high frequency had a boundary problem at .frequency = 1907. Fix by using <= comparison instead of <.

• Add some bounds checking to make sure values fit in 16 bits.

• Unrelated, missed getting into a previous PR: Remove extraneous include in shared-bindings/adafruit_bus_device/SPIDevice.c.

@jedgarpark Could you test this after the artifact builds? It's more complicated than the version I gave you earlier to test.

[jedgarpark]

Tested it, this works perfectly now, thank you @dhalbert

[tannewt]

I've gotta look at this further when I'm fresh. See section 4.5.2.2 in the datasheet. My understanding was that CC needs to be TOP+1 in order to not go low at all.

[dhalbert]

I see what you mean in the datasheet. I will look at the output with the Saleae. We may need to use 65534 as the TOP to set instead of 65535.

[dhalbert]

@tannewt I read the datasheet and did a lot of boundary testing with the Saleae, checking both the low and high frequency cases at the boundary , and much higher and lower frequencies. I made sure duty_cycle = 65535 result in high all the time, no glitches output. Just rounding properly did not always result in no glitches. I also improved the calculations of actual_frequency by rounding to the nearest integer.

[tannewt]

Ok great! Thanks @dhalbert