match return type inconsistent with 5.x semantics and actual return type

[joshgummersall]

I just noticed something strange around the top-level match method.

In version 5.1.0, the match function return type is a union of the return types of the case functions:

In version 6.3.0, the match function return type is an array of the return types of the case functions:

When executed, the return type appears to always be consistent with the 5.1.0 return type above.

[joshgummersall]

I also just noticed that parameter types are not properly inferred for case functions.

In 6.3.0 you get:

In 5.1.0 you get:

[yuhr]

Hmm, the return type can be fixed easily, but the latter problem (parameter types) seems quite difficult to be resolved, possibly due to inability of TS's type inference, and thus it might need regression to the old implementation before #166, which is very annoying... I'm going to focus on seeking another way around this.