12$: Optimize SIMD bitset iteration

[strager]

Block comment parsing uses a SIMD register. It iterates over each true in the register:

https://github.com/quick-lint/quick-lint-js/blob/bd583a1b34bfa557c99477551af68525d8e77cc2/src/quick-lint-js/fe/lex.cpp#L1942-L1945

We can adapt Kernighan's popcount algorithm to speed up this up: https://www.techiedelight.com/brian-kernighans-algorithm-count-set-bits-integer/

[strager]

Roughly:

1. If the mask is 0, exit.
2. Use countr_zero to locate the right-most 1.
3. Yield the bit.
4. Update the mask: mask = mask & (mask - 1);
5. Go to step 1.

(and we'd need to keep track of the index)