tests/c_lib: Skip double sqrt test when double is not 64 bits

This test only works for 64-bit values. On targets which use 32 bits for
double, skip this version. The float test will check the 32-bit math for
that.

This is the 'nicer' version of this fix which uses sizeof rather than
relying on __SIZEOF_DOUBLE__.

Signed-off-by: Keith Packard <keithp@keithp.com>