A more useful Monoid instance for Data.Map

[Gabriel Gonzalez]

-1 on breaking the existing Monoid instance.  I'd prefer either:

a) A newtype for the additional Monoid instance, or
b) providing an associative operator equivalent to the new Monoid instance

I'm not convinced that the replacement instance is more useful, let 
alone useful enough to break everything depending on containers.  A lot 
of packages both on and off Hackage depend on Edward's code, almost all 
of which would break as a result of this change.