Rationalise Error classes

[scott-griffiths]

I can't remember the motivation for having so many specialised error classes (I suspect just because I could) and they can become a bit annoying.

Sometimes a given issue should be an InterpretError, and sometimes a CreationError depending on the context the code is used in. Both derive from ValueError, and I think I'd be happy for them both to be ValueErrors.

Similarly ReadError derives from IndexError and doesn't add anything to it.

ByteAlignError is at least a bit distinctive, and gives slightly more information than alternatives.

Suggest that for version 4.2 we change from deriving from the base class to just being an alias for it. That shouldn't break any code and would allow some simplifications