drop the PartialEq impl #178
Description
I'm now of the opinion that merging #81 was a mistake. I think having a PartialEq impl on Regex is incredibly misleading, even though it is clearly documented. In particular, regular expressions do actually have a formal notion of equality, and it is quite expensive to compute. We simply shouldn't expose any PartialEq or Eq impls.
If users need a PartialEq impl (e.g., to store regexes as keys in a HashMap?), then I think requiring a newtype is fine.