Even if checking for zero wasn't the most efficient way to implement it, a
compiler is allowed to transform

> if (divisor == 0) [[unlikely]] throw ...

into something more optimal. Maybe there is a way to handle the hardware
trap on division and recover from that with an exception throw. The cost of
a division is also relatively large, and I doubt that the zero-check has
much performance impact.

Anyhow, any proposal will obviously need some feedback from implementers,
performance measurements, etc. Until proven otherwise, checking for zero is
a perfectly valid solution already, and adding such a "div_with_exception"
function would be a waste of committee time.