Optimize Zen 4 GEMM macro block sizes (P, Q, R)

[Vasudeva-bit]

Zen 4 Cache Boundary Optimisations (uses CooperLake Micro Kernels) #5837

Description:
This PR introduces targeted cache geometry overrides (P, Q, R parameters) for the Zen 4 architecture. Parameters were derived using Bayesian Optimization (Optuna) to perfectly align matrix panel footprints with the physical 1MB L2 and 32MB L3 limits of the Zen 4 CCX.

1. Successful Optimisations (Implemented)

These standard GEMM variants demonstrated absolute GFLOPs scaling over the baseline OpenBLAS generic fallbacks. These are the configurations included in this patch.

Variant	OpenBLAS Baseline (GFLOPs)	BO Tuned Parameters (P, Q, R)	BO Tuned Performance (GFLOPs)	Net Improvement
SGEMM	2880.00	384, 512, 5936	3080.71	+ 6.9%
DGEMM	1293.73	512, 512, 2288	1511.87	+ 16.8%
CGEMM	2981.33	160, 480, 528	3143.95	+ 5.4%
ZGEMM	1288.60	176, 256, 1520	1459.45	+ 13.2%

2. Discarded

Variant	OpenBLAS Baseline (GFLOPs)	BO Tuned Parameters (P, Q, R)	BO Tuned Performance (GFLOPs)
SBGEMM	5318.71	368, 712, 4544	5267.25
CGEMM_3M	2606.98	168, 376, 3920	2655.71
ZGEMM_3M	960.02	192, 256, 1904	889.82