Should we get rid of nb::Result for inherent functions?

[bjoernQ]

Should be a general decision - not just for SPI.
If we don't want that - what about embedded-hal-nb?

[jessebraham]

I think probably it's reasonable to return an Option from read_byte based on whether or not we're able to read, then we can just converted this into the appropriate return type in the embedded_hal_nb implementation (e.g. Some -> Ok, None -> Err).

[bjoernQ]

Had another look and while for something like read_byte returning Option even feels more natural there are other places where it's not that easy.

e.g. write_byte,flush_tx, some functions in the crypto-accelerator drivers (some are questionable - e.g. Hmac::configure never returns WouldBlock), there are usages in SPI and TWAI, in PeriodicTimer::wait and ADC