Description:
------------
Current implementation of BitSet::cardinality() counts bits one at a time. This 
can be done a lot better using known fast methods for computing the "hamming 
weight". The easiest way (for Linux) would be to check if the compiler is GCC and 
use the __builtin_popcount() it provides, or use one of the implementations shown 
at http://en.wikipedia.org/wiki/Hamming_weight.

Probably it would also need to change the type of bitset_val on 
https://github.com/php/pecl-numbers-bitset/blob/master/php_bitset.h from unsigned 
char* to unsigned long* (or unsigned long long * etc, so 32/64 bits are handled 
at a time).