Fix nRF UART reset

[tannewt]

disable only turns off ENABLE but doesn't set the init tracking that
nrfx uses. uninit hangs if ENABLE is off and is called because it
waits forever for TX to stop.

[dhalbert]

SPI and I2C do it a different way, like this, in the constructor. They do disable in deinit(), and in the constructor, they try to do do an nrfx init(), and only if that fails do they do the uninit(). I can't remember exactly why, but this was a better choice at the time. Is UARTE different for some reason? Maybe it should follow the pattern in SPI and I2C.

    nrfx_err_t err = nrfx_spim_init(&self->spim_peripheral->spim, &config, NULL, NULL);

    // A soft reset doesn't uninit the driver so we might end up with a invalid state
    if (err == NRFX_ERROR_INVALID_STATE) {
        nrfx_spim_uninit(&self->spim_peripheral->spim);
        err = nrfx_spim_init(&self->spim_peripheral->spim, &config, NULL, NULL);
    }

[tannewt]

I think the I2C and SPI pattern is an artifact of the early code that didn't correctly reset the peripherals on soft reset. Want me to update the other two as well?

It looks like the I2C disable was added here: https://github.com/adafruit/circuitpython/pull/1277/files#diff-0fadf6b9ea8a38a03206d10c865959c3R59

[dhalbert]

OK, that makes sense to me. The point is that nrfx_..._uninit() is now idempotent in our fork of nrfx. I guess I changed that but didn't go back and utilize the new capability. So yes, please fix the other busio impls. Thank you.

[tannewt]

@dhalbert Done!

[dhalbert]

Excellent! Thank you for clear thinking on all of this, and debugging the neopixel PWM issue.