While I was working on the code base of embedded-io, I noticed some behavioral divergences between the default implementation of ``std::io::Read::read_exactandembedded-io::Read::read_exact`. The first retries to read the source every time that an error with an `ErrorKind::Interrupted` kind is caught, while the latter fails as soon as a `ErrorKind::Interrupted` is yield by the source.
Is this intentional?
TBH, I am a bit surprised by the std::io choice because read doesn't retry to read.
This creates some inconsistencies between read and read_exact.