> The problem with having three ways of declaring a variable

Three ways? I only count two, and the rule for what to use is very simple: any time the type can be inferred from the initial value, you use := (obviously you can't use := in other cases, so there isn't really any ambiguity).