<cmath>: Tolerate 80-bit long double

[statementreply]

Use the same code for long double lerp overloads as other math functions. This avoids calling long double _Common_lerp, which inspects long double's bit representation.

[StephanTLavavej]

I don't care about adding support for 80-bit long double where it didn't previously work, but this increases consistency with 3-arg hypot etc. and is therefore worthwhile. Thanks!