gh-132509: Speed up bytes iteration in FOR_ITER_VIRTUAL

[Wulian233]

Benchmark	Main	PR	Throughput improvement
Empty loop	13.858 ns/item	11.850 ns/item	16.9% faster
XOR accumulation	28.195 ns/item	25.409 ns/item	11.0% faster
Conditional count	28.649 ns/item	24.721 ns/item	15.9% faster

Details

test script

import statistics
import timeit

DATA = bytes(range(256)) * 4096  # 1 MiB
NUMBER = 10
REPEAT = 11


def loop_pass():
    for byte in DATA:
        pass


def loop_xor():
    result = 0
    for byte in DATA:
        result ^= byte
    return result


def loop_count():
    result = 0
    for byte in DATA:
        if byte == 255:
            result += 1
    return result


BENCHMARKS = (
    ("empty loop", loop_pass),
    ("xor accumulation", loop_xor),
    ("conditional count", loop_count),
)

# Warm up adaptive specialization.
for _, benchmark in BENCHMARKS:
    for _ in range(10):
        benchmark()

print("| Benchmark | Median | Minimum | Median ns/item |")
print("|---|---:|---:|---:|")

for name, benchmark in BENCHMARKS:
    samples = timeit.repeat(
        benchmark,
        number=NUMBER,
        repeat=REPEAT,
    )
    median = statistics.median(samples)
    minimum = min(samples)
    ns_per_item = median * 1e9 / (NUMBER * len(DATA))
    print(
        f"| {name} | {median:.9f} s | "
        f"{minimum:.9f} s | {ns_per_item:.3f} |"
    )

• Issue: Use tagged pointers for more efficient integer operations. #132509