[mlir][Aarch64] Add lowering support from unsigned `vector.contract` to i8mm

[dcaballe]

LowerContractionToSMMLAPattern.cpp added support to lower a vector.contract to i8mm's smmla instructions. However, the existing code can only handle signed i8,i8,i32 matrix multiplication cases. We should generalize (and rename) the pass to also handle unsigned i8,i8,i32 cases by adding matching support for extui and generating ummla instructions instead.

#Gemma