Add KIND contant to Interface trait

[rfuest]

This PR adds the interface kind constant to the Interface trait as discussed in #167.

I've decided to not extend the InitError enum and just panic instead. It should only be possible to encounter this error during development time, when the programmer wasn't aware that the combination of controller model, color format, and interface kind isn't supported.

[almindor]

LGTM!

[almindor]

Hmm actually thinking about this, I wonder if forcing panic is the wrong call here. I think I'd sleep better if we Model::init errored out properly in this case.

I'm not sure how to best approach that given how Interface::Error is currently generic. My main concern is that kernels can have basic expectations such as "drivers do not panic" and I think that should be upheld, even in init code.

[rfuest]

Hmm actually thinking about this, I wonder if forcing panic is the wrong call here. I think I'd sleep better if we Model::init errored out properly in this case.

I'm not sure how to best approach that given how Interface::Error is currently generic. My main concern is that kernels can have basic expectations such as "drivers do not panic" and I think that should be upheld, even in init code.

I don't think that this would have been an issue, because the compiler would have been able to resolve the assertions at compile time and remove them from the final binary.

I've changed the code to return a ModelInitError from Model::init, which is a variant of InitError that doesn't contain the reset pin error. Using InitError directly would have required adding an additional type parameter for the reset pin error to Model::init. This didn't feel right because Model::init has no access to the reset pin.

If you are concerned about possible panics we should also look at Builder::display_size, which is much more likely to panic IMO, because the parameters might not be known at compile time. We could move the size check to Builder::init and also return a InitError::InvalidConfiguration in that case. But this is outside the scope of this PR.

[almindor]

If you are concerned about possible panics we should also look at Builder::display_size, which is much more likely to panic IMO, because the parameters might not be known at compile time. We could move the size check to Builder::init and also return a InitError::InvalidConfiguration in that case. But this is outside the scope of this PR.

Agreed, I'll make an issue for that