WIP: DsExpr: Make result of mfix in rec do more lazy
With -XStrict on this result is matched strictly which is probably not desired. There doesn't seem to be a way for the programmer to make this match irrefutable when XStrict is on.
Alternatives may be looking at any lazyness annotations on the bound variables and "bubbling" those up to the tuple match.
I'm not 100% certain about the second LazyPat
being used here. Although it
enables the mdorun007 test to pass, I suspect that this may be due to some
fortunate interaction with the strictness analyzer rather than this being the
proper fix.