refactor(compute): extract shared matmul helpers to gpu_engine_matmul.go

Consolidate repeated patterns from 14 quantized matmul methods into 6
shared helpers: uploadRawBytes, aShapeCheck2D, bweightShapeMKN,
quantGemvResult, dequantSgemm, and sgemmNTOrFallback. Each original
method is now a thin wrapper calling these helpers, reducing
gpu_engine.go by 797 lines (net -557 across both files).

Zero behavioral changes -- all method signatures remain identical.

Author: dndungu