scalbnf should wrap ldexpf

[ZERICO2005]

float scalbnf(float x, int exp) should wrap float ldexpf(float x, int exp) instead of doing x * exp2f((float)exp). These functions are equivalent because FLT_RADIX == 2.

This could be fixed by having scalbnf as an alias to ldexpf. Similar thing for scalblnf(float x, long exp) apart from casting long to int.