Add random to ESP32-S2, fix it on STM32

[hierophect]

This PR adds the TRNG to ESP32-S2, based off the cxd56 implementation, which has a similar system call. This will enable use of os.random().

It also fixes a serious but overlooked issue with the STM32 os.random call where it would not supply values past the first 8 bit number. (Yikes!)

[tannewt]

Just a couple questions.

[anecdata]

I don't know what the right answer is here, but pseudo-random is fine for many purposes other than cryptographic. Users can also seed with their own entropy. It would be a shame to disable random completely when wi-fi is not up. Maybe an alert in the docs would suffice.

[hierophect]

I don't know what the right answer is here, but pseudo-random is fine for many purposes other than cryptographic. Users can also seed with their own entropy. It would be a shame to disable random completely when wi-fi is not up. Maybe an alert in the docs would suffice.

Circuitpython's built in random module is already pseudorandom, and should activate automatically when os.urandom is not available. But os.urandom itself is supposed to be TRNG (True Random Number Generation), and is represented as such in the documentation - I don't think we should provide pseudorandom numbers while advertising them as TRNG.

[hierophect]

@anecdata are you using the random module? You should still be able to access pseudorandom numbers without this specific OS feature. I may have mistaken CIRCUITPY_RANDOM for standing in for the TRNG when it was actually a flag for both. If that's the case, once it's re-enabled you'll be able to get pseudorandom values even when TRNG is turned off.

[anecdata]

Sorry, yes, I conflated the two since the random module went away from esp32-s2 in a recent build, and I don't think os.urandom() was present either. So I think it will be all good. Thanks.

[hierophect]

Sorry, yes, I conflated the two since the random module went away from esp32-s2 in a recent build, and I don't think os.urandom() was present either. So I think it will be all good. Thanks.

That was probably my fault, I removed CIRCUITPY_RANDOM because I thought it was the flag for the TRNG only. It must have also disabled the pseudorandom numbers, which I didn't intend.

[dhalbert]

Merge conflicts here. Is this otherwise good to go? @tannewt did you have anything further?

[dhalbert]

Looks like it's all set.

addressed

[hierophect]

Doesn't look like I can merge until @tannewt signs off.

[hierophect]

nevermind