Floating-point literal types problem

[amishne]

Unlike integers or null, where the literal is type-less, floating-point literals do have specific types. This needs to be used to enhance the type validation, which rights now treats them too leniently.

[amishne]

0.1e2-style notation is applicable for halfs, floats and doubles. All other types must appear in their appropriate hex format only. However, the situation right now is that every floating-point literal is legal for any floating-point type.