[lint request] unnecessary_float_transmute

[gnzlbg]

The f32 and f64 types have pub fn from_bits(u32) -> f32 and pub fn to_bits(f32) -> u32 methods, yet many users don't know about them, and use transmute instead.

We should add a lint unnecessary_float_transmute that:

• detects a transmute between a floating-point type and the unsigned integer type of the same bit width, and recommends using either from_bits or to_bits instead.

• detects a transmute between a floating-point type and the signed integer type of the same bit-diwth, and recommends using from_bits/to_bits + an as cast (from unsigned to signed or vice-versa) instead.

When one of the arguments to the transmute is a macro variable, this lint probably can't work correctly (e.g. the macro might be expanded with types that aren't floats), so the lint should bail.

[Centril]

I think this may also be a candidate for rustc uplifting.

[krishna-veerareddy]

@gnzlbg Hey looks like the lint for int to float conversion already exists. I can tackle the the float to int conversion.