Improve emulation of AVX2 min/max 64-bit

[r-devulap]

Improves benchmarks by about 1.25x.

Benchmark                                                                   Time             CPU      Time Old      Time New       CPU Old       CPU New
--------------------------------------------------------------------------------------------------------------------------------------------------------
[simdsort/random_.*int64_t vs. simdsort/random_.*int64_t]                -0.1597         -0.1595         61073         51322         61080         51340
[simdsort/random_.*int64_t vs. simdsort/random_.*int64_t]                -0.1482         -0.1482       1926998       1641449       1926924       1641282
[simdsort/random_.*int64_t vs. simdsort/random_.*int64_t]                -0.1385         -0.1386      23119151      19916947      23117303      19914147
[simdsort/random_.*int64_t vs. simdsort/random_.*int64_t]                -0.1305         -0.1306     274425664     238612093     274388498     238561940
[simdsort/random_.*int64_t vs. simdsort/random_.*int64_t]                -0.1942         -0.1943         54594         43993         54612         44002
[simdsort/random_.*int64_t vs. simdsort/random_.*int64_t]                -0.1528         -0.1529       1699531       1439793       1699414       1439623
[simdsort/random_.*int64_t vs. simdsort/random_.*int64_t]                -0.1404         -0.1404      20536591      17654205      20534818      17652359
[simdsort/random_.*int64_t vs. simdsort/random_.*int64_t]                -0.1287         -0.1287     244993708     213467588     244956884     213438188