Use SSE2 in the ASCII fast path of decodeUtf8

[ethercrow]

Before:

benchmarking DecodeUtf8/Strict+ascii
time                 18.06 ms   (17.12 ms .. 18.56 ms)
                     0.994 R²   (0.984 R² .. 1.000 R²)
mean                 19.66 ms   (18.57 ms .. 24.01 ms)
std dev              4.971 ms   (25.90 μs .. 9.169 ms)
variance introduced by outliers: 86% (severely inflated)

benchmarking DecodeUtf8/Stream+ascii
time                 19.11 ms   (19.08 ms .. 19.13 ms)
                     1.000 R²   (1.000 R² .. 1.000 R²)
mean                 19.16 ms   (19.13 ms .. 19.22 ms)
std dev              100.5 μs   (26.19 μs .. 160.2 μs)

benchmarking DecodeUtf8/Lazy+ascii
time                 18.78 ms   (18.75 ms .. 18.80 ms)
                     1.000 R²   (1.000 R² .. 1.000 R²)
mean                 18.83 ms   (18.80 ms .. 18.88 ms)
std dev              92.52 μs   (25.88 μs .. 141.1 μs)

After:

benchmarking DecodeUtf8/Strict+ascii
time                 11.17 ms   (10.62 ms .. 11.56 ms)
                     0.992 R²   (0.978 R² .. 0.999 R²)
mean                 11.88 ms   (11.07 ms .. 15.03 ms)
std dev              3.993 ms   (305.1 μs .. 8.201 ms)
variance introduced by outliers: 92% (severely inflated)

benchmarking DecodeUtf8/Stream+ascii
time                 6.647 ms   (6.620 ms .. 6.673 ms)
                     1.000 R²   (1.000 R² .. 1.000 R²)
mean                 6.709 ms   (6.679 ms .. 6.809 ms)
std dev              137.3 μs   (37.82 μs .. 276.3 μs)

benchmarking DecodeUtf8/Lazy+ascii
time                 6.421 ms   (6.372 ms .. 6.479 ms)
                     1.000 R²   (0.999 R² .. 1.000 R²)
mean                 6.447 ms   (6.423 ms .. 6.527 ms)
std dev              123.4 μs   (44.11 μs .. 246.1 μs)