Allow uses of Try with nullable error types (Try<S, F?>)

[qnga]

No description provided.

[mickael-menu]

I changed my mind. In which case does it make sense to have a nullable failure type instead of a null result?

[qnga]

To convey the idea that there was a failure we know nothing about. Anyway, there's a mistake in the current code: this kind of types is accepted by the compiler but fails at runtime.

[mickael-menu]

Fair enough for the runtime failure. I'm not sure I would use the nullable failure even in the case you mentioned though. I guess we'll see on a case-by-case basis.