> Especially for floating-point, where the distinction between a < b and a <= b is not straightforward.

I would argue that if your floating point code cares about < vs <= then you are probably making a mistake.