pub const NEG_FLOAT: u8 = 0x1F;Expand description
Typed unary - on a Float operand. Same shape as NEG_INT
but on the float side: raw f64::from_bits, IEEE 754 negation
(sign-bit flip), push as NanValue::new_float. Skips the
NEG tag-dispatch chain and preserves -0.0 because the
negation runs on the float bit pattern, not on 0 - x desugar.