[CPP] Add solution 6: 1-bit Wheel 8-of-30 Sieve

[bonnhatnguyen]

Description

This pull request adds PrimeCPP/solution_6, introducing an ultra-fast, faithful, 1-bit Wheel 8-of-30 Sieve of Eratosthenes implementation in C++17.

Algorithmic Highlights

• Wheel 8-of-30 Factorization: Skips multiples of 2, 3, and 5. For every 30 numbers, only 8 coprime candidates are traversed ([1, 7, 11, 13, 17, 19, 23, 29]).
• Register-Resident 8-Stride Cycle: The 8 strides ([3, 2, 1, 2, 1, 2, 3, 1] * factor) fit entirely into CPU registers (rax, rbx, rcx, rdx, r8..r11). The inner loop is fully unrolled by the 8-stride cycle, requiring zero memory lookups, zero modulo, and zero branch mispredictions during sieve marking.
• Strict L1 Data Cache Residency: Compact 1-bit representation requires ~32.55 KB, fitting 100% inside CPU L1 Data Cache (32 KB - 48 KB) to eliminate L2/L3 cache latency.
• Strict Faithfulness: Dynamic runtime allocation per sieve pass, full class encapsulation, and verified against 78,498 primes up to 1,000,000.
• Hardware-Concurrency Parallelism: Seamless multi-threaded scaling across all logical cores.

Benchmark Results (Intel Core i7-12700H, Docker Ubuntu 22.04)

• Single-threaded: ~8,067 - 8,672 passes/sec (40,336+ passes in 5.00s) — outperforming existing faithful C++ solutions (~10x faster than Solution 1).
• Multi-threaded (20 threads): ~41,618 - 43,112 passes/sec (208,315+ passes in 5.00s).

ndt0208-cpp-wheel8;40336;5.00021;1;algorithm=wheel,faithful=yes,bits=1
ndt0208-cpp-wheel8-par;208315;5.00534;20;algorithm=wheel,faithful=yes,bits=1

Contributing requirements

• I read the contribution guidelines in CONTRIBUTING.md.
• I placed my solution in the correct solution folder.
• I added a README.md with the right badge(s).
• I added a Dockerfile that builds and runs my solution.
• I selected drag-race as the target branch.
• All code herein is licensed compatible with BSD-3.