Support: Sparse matrix optimisations in IREE

[shreya-um]

I’m interested in exploring sparse matrix optimization and support in IREE.
Specifically, I’d like to understand what’s currently available at the MLIR level e.g., in the linalg or sparse_tensor abstractions for representing or optimizing sparse computations.
(and not Vulkan backend-specific optimizations, but rather to the generic CPU path and the compiler-level transformations)