Special cases aside, I think this is a YAGNI. The "obvious" formulas `(a + b)/2)` and `0.5 * (a + b)` _do_ do exactly the right thing (including giving a perfectly correctly-rounded answer with round-ties-to-even on a typical IEEE 754-using machine) provided that subnormals and values very close to the upper limit are avoided. If you're doing floating-point arithmetic with values of size > 1e300, you've probably already got significant issues.

I could see specialist uses for this, e.g., in a general purpose bisection algorithm, but I'm not convinced it's worth adding something to the math library just for that.