#include <float.h> in avutil-eval.c.

The code uses DBL_MAX.  Apparently it's expecting <limits.h> to bring
that in, but C99 says that <limits.h> is for _integer_ type limits; for
floating-point limits, it's <float.h>.