Thank you for this neat crate, especially with the good test coverage and the typestate implementation.

As the enable/disable functions are consuming, the struct including the pins is dropped:

However, this can be a problem for some systems that operate a singleton principle for GPIO pins. On such systems, it is only possible to obtain a pin instance once. An example of this is the RP2040 HAL:

It is not possible for users to create new instances of a Pin. Singleton instances of each pin are made available to users through the Pins struct.

Of course, there are workaround (e.g. wrapped pins with internal reference), but that would be very inconvenient and not elegant.

I see two possible solutions in case of an error:

1. Releasing the internal pins as part of the Error enum
2. Return a tuple containing the error + struct with the old state

Personally, I tend towards the second solution
What do you think?

Looking forward to your feedback.

PS: If desired, I can also support with an appropriate PR.