Update name_to_dnswire_loop

[k0ekk0ek]

This updates my take on the function and gives it a slight boost. It's still somewhat slower for the top-1m.csv, but faster for just benchmark or a csv of all owners in the .se zone. What's interesting is that the prefix-minimum is quite a bit slower than any of the simdjson-like approaches for both benchmark and .se.

$ ./benchmark top-1m.csv
loaded 1000000 names
average length 25.5398 bytes/name
string with lengths between 0 and 31 : 770551 (77.06 %)
string with lengths between 32 and 63 : 203682 (20.37 %)
string with lengths between 64 and 95 : 22182 (2.22 %)
string with lengths between 96 and 127 : 1622 (0.16 %)
string with lengths between 128 and 159 : 517 (0.05 %)
string with lengths between 160 and 191 : 490 (0.05 %)
string with lengths between 192 and 223 : 464 (0.05 %)
string with lengths between 224 and 255 : 492 (0.05 %)

name_to_dnswire_idx_avx is simdjson-like
name_to_dnswire_avx is Prefix-Minimum
name_to_dnswire is conventional

name_to_dnswire_idx_avx512     :   2.01 GB/s   78.6 Ma/s  12.72 ns/d   3.79 GHz  48.15 c/d  65.33 i/d    1.9 c/b   2.56 i/b   1.36 i/c 
name_to_dnswire_idx_avx        :   1.98 GB/s   77.5 Ma/s  12.90 ns/d   3.87 GHz  49.87 c/d  80.71 i/d    2.0 c/b   3.16 i/b   1.62 i/c 
name_to_dnswire_loop           :   1.74 GB/s   68.2 Ma/s  14.67 ns/d   3.87 GHz  56.73 c/d  87.07 i/d    2.2 c/b   3.41 i/b   1.53 i/c 
name_to_dnswire_avx            :   2.21 GB/s   86.7 Ma/s  11.53 ns/d   3.86 GHz  44.54 c/d  80.14 i/d    1.7 c/b   3.14 i/b   1.80 i/c 
name_to_dnswire_simd           :   2.03 GB/s   79.5 Ma/s  12.58 ns/d   3.86 GHz  48.59 c/d  113.62 i/d    1.9 c/b   4.45 i/b   2.34 i/c 
name_to_dnswire_scalar_labels  :   1.33 GB/s   52.0 Ma/s  19.22 ns/d   3.86 GHz  74.27 c/d  119.43 i/d    2.9 c/b   4.68 i/b   1.61 i/c 
name_to_dnswire                :   0.63 GB/s   24.5 Ma/s  40.83 ns/d   3.83 GHz  156.45 c/d  258.27 i/d    6.1 c/b  10.11 i/b   1.65 i/c