[Dr-Emann]

Add count_bits for computing the popcount of a full bitmap store, used by BitmapStore::try_from and internal_validate. It uses a native vector popcount on NEON and AVX-512, and a Harley-Seal carry-save adder popcount on other targets, similar to CRoaring.

group                                                         main                                   simd
-----                                                         ----                                   ----
Successive Or/Multi Or Owned/census-income                    1.04    733.0±4.58µs        ? ?/sec    1.00    707.8±5.19µs        ? ?/sec
Successive Or/Multi Or Owned/weather_sept_85                  1.00      3.4±0.02ms        ? ?/sec    1.00      3.4±0.02ms        ? ?/sec
Successive Or/Multi Or Ref/census-income                      1.04   724.9±62.14µs        ? ?/sec    1.00    695.8±7.22µs        ? ?/sec
Successive Or/Multi Or Ref/weather_sept_85                    1.01      3.4±0.03ms        ? ?/sec    1.00      3.4±0.02ms        ? ?/sec
Successive Or/Successive Or Assign Owned/census-income        1.01   979.7±12.99µs        ? ?/sec    1.00    968.9±5.67µs        ? ?/sec
Successive Or/Successive Or Assign Owned/weather_sept_85      1.00      3.4±0.03ms        ? ?/sec    1.00      3.4±0.02ms        ? ?/sec
Successive Or/Successive Or Assign Ref/census-income          1.01   959.4±12.58µs        ? ?/sec    1.00    954.6±4.65µs        ? ?/sec
Successive Or/Successive Or Assign Ref/weather_sept_85        1.01      3.4±0.05ms        ? ?/sec    1.00      3.4±0.02ms        ? ?/sec
Successive Or/Successive Or Ref Ref/census-income             1.00  1066.5±13.54µs        ? ?/sec    1.00   1062.7±8.71µs        ? ?/sec
Successive Or/Successive Or Ref Ref/weather_sept_85           1.00      4.1±0.05ms        ? ?/sec    1.00      4.1±0.03ms        ? ?/sec
deserialization/deserialize_from/census-income                1.02    439.1±3.07µs     5.0 GB/sec    1.00    431.6±7.17µs     5.1 GB/sec
deserialization/deserialize_from/weather_sept_85              1.02  1859.2±21.77µs     4.4 GB/sec    1.00  1831.4±12.78µs     4.5 GB/sec
deserialization/deserialize_unchecked_from/census-income      1.07    131.2±2.52µs    16.8 GB/sec    1.00    123.1±2.48µs    17.9 GB/sec
deserialization/deserialize_unchecked_from/weather_sept_85    1.00    422.4±5.75µs    19.3 GB/sec    1.00    421.5±4.87µs    19.3 GB/sec
pairwise_and/assign_ref/census-income                         1.01   609.0±40.62µs        ? ?/sec    1.00   604.9±34.98µs        ? ?/sec
pairwise_and/assign_ref/weather_sept_85                       1.02      2.3±0.10ms        ? ?/sec    1.00      2.2±0.08ms        ? ?/sec
pairwise_or/assign_ref/census-income                          1.01  1075.1±35.94µs        ? ?/sec    1.00  1067.2±30.78µs        ? ?/sec
pairwise_or/assign_ref/weather_sept_85                        1.02      5.6±0.16ms        ? ?/sec    1.00      5.4±0.14ms        ? ?/sec
pairwise_sub/assign_ref/census-income                         1.01   684.3±31.10µs        ? ?/sec    1.00   679.9±35.94µs        ? ?/sec
pairwise_sub/assign_ref/weather_sept_85                       1.01      2.9±0.10ms        ? ?/sec    1.00      2.8±0.12ms        ? ?/sec
pairwise_xor/assign_ref/census-income                         1.01  1095.5±37.79µs        ? ?/sec    1.00  1087.3±25.82µs        ? ?/sec
pairwise_xor/assign_ref/weather_sept_85                       1.01      5.5±0.11ms        ? ?/sec    1.00      5.4±0.14ms        ? ?/sec

Only a few percent faster, but seemingly consistently faster, on my M1 Pro, at least.