Add frequencyio.FrequencyIn

[sommersoft]

Its time again to try and get this in. Updates since #1144:

• Instead of putting this in pulseio, this creates a new frequencyio to better manage inclusion/exclusion. Most of the smaller builds (M0 basic, CPX, etc) were exceeding flash size.

• Now uses a single "reference TC" for interrupts/results calculation, and "capture TCs" for frequency sensing. This keeps the interrupts from flooding the MCU, but does eliminate 1 timer available for capture. I've had up to 2MHz input detected with no detrimental effects to the VM, USB, or REPL.

• Since the "capture TCs" now use a simple count-of-events based on a time period, I added a capture_period property. Notes about this are in the documentation. There is some math-assumption built in to the frequency calculation, since the reference TC will not catch the time periods exactly.

• I'm getting a variance of 10% maximum on readings. Using a sample average has proven the best way to smooth this out. Here are some example Mu screenshots with a 2KHz signal:

To anyone testing/reviewing: please, don't be gentle with it! Break it. Criticize (constructively)! All of that is appreciated on my end. 😄

NOTE: I have left the previous SAMD21 code in, even though it isn't included by default. I still need to go back through and test it with a custom build, but I would like to leave it as an available option.

[dhalbert]

This looks good! In the long run maybe the DPLL1 stuff might be factored out into samd-peripherals, in case other code wants to do similar stuff, but it's fine for now. This has been a long road, but it's also been tested a lot as a result.

[sommersoft]

In the long run maybe the DPLL1 stuff might be factored out into samd-peripherals, in case other code wants to do similar stuff

Yeah, that is totally doable I think. I can work that soon.