bench: execute publication-grade controlled-hardware measurements

[sauravsingla]

Goal

Execute the controlled-hardware campaign defined by Issue #9 once suitable dedicated hardware is available.

Blocked by

• Issue bench: define publication-grade controlled-hardware campaign #9 publication-grade execution contract
• access to dedicated Linux CPU hardware with genuine NUMA/multi-socket topology for the NUMA portion
• resolved immutable dataset and competitor pins

Required measurements

• 1/2/4/8/16/32+ thread scaling where supported
• genuine NUMA placement/locality measurements
• hardware counters and ablations
• same-hardware VeloGraphX vs pinned native LAGraph/GraphBLAS and GAP comparisons
• repeated public-dataset crossover measurements
• validated raw and summary artifacts

Claim rule

No publication-grade performance or scalability claims should be made until the readiness gate defined in Issue #9 passes and these measurements are complete.