Remove pointless gpio attach_interrupt overrides

[peterbarker]

This might be contentious. But having all of these empty definitions - which returned true for some strange reason - was painful when I was playing with the attach-a-functor change.

[peterbarker]

Just two questions:

• are we good with having a base-class implementation rather than requiring all HALs to override
• are we good with changing the behaviour on all of these so they stop pretending they successfully attached an interrupt?

[peterbarker]

... amongst our questions are such diverse elements as "what's up with std::swap, anyway?"

[WickedShell]

• are we good with having a base-class implementation rather than requiring all HALs to override

Sounds reasonable to me, less reimplemented/cruft in the HAL's is a win as far as I can tell.

• are we good with changing the behaviour on all of these so they stop pretending they successfully attached an interrupt?

Absolutely. That's a terrible behavior to just always return true their, defeats the API, and makes it hard for the caller to safely/sanely handle it.

[peterbarker]

It was decided at the devcall that std::swap was OK so long as godbolt approves of it.

The HAL behaviour re: attaching interrupts currently sucks, so fix it.

pure-virtual isn't an attribute we need in the HAL interface, so kill the pointless overload.

[peterbarker]

https://godbolt.org/z/Tk0tqS <- std::swap looks OK.