Add lint that detects transmutation from a float to an integer and suggests usage of `{f32, f64}.to_bits()` instead.