feat: add Ascend hardware benchmarks

[baominghelly]

Summary

• add the native Ascend memory bandwidth, STREAM, and ACL D2D memcpy size sweep benchmark suite
• integrate Ascend platform aliases, runtime detection, build selection, and result parsing into the existing hardware adapter
• add focused adapter tests and update hardware benchmark documentation

Testing

• full test suite: 42 passed
• platform-focused tests: 40 passed
• Black 23.9.1: passed
• Python compileall: passed
• shell syntax checks: passed
• git diff --check: passed
• Ascend 910B4 physical device 2: native build passed
• Ascend 910B4 physical device 2: native npu_perf_suite --all passed
• Ascend 910B4 physical device 2: public hardware.cudaUnified.Comprehensive adapter path passed
• validated exactly 9 expected metrics: 4 bandwidth sweep, 4 STREAM, and 1 ACL D2D memcpy size sweep
• current head: 6d1b239afcf4eca68db218893b3076e0a55aa33f
• tested source tree: 24929d013951298ac8d814a903c928d6fbe9a1d8 (the current head resolves to this exact tree)