Make sealed generic function redefinable

[marcoheisig]

While technically a violation of the protocol, I think it would be useful if one could redefine the sealed methods of a sealed generic function. At least, it would be useful for people developing a library that uses sealed generic functions.

Of course, doing so brings the same problems as redefining constants or inlined functions. I think the best approach is to signal a 'redefinition-of-sealed-metaobject' error and to provide an appropriate 'continue' restart.