hex-geglu

This release introduces significant enhancements to the Hexagon backend, including multi-NPU device support, a fully asynchronous backend, and improved performance for various operations like ALLREDUCE and matrix multiplications. It also includes numerous bug fixes and optimizations across different Hexagon modules.

This release focuses on optimizations and internal refactoring for hexagon activation operations, including improved microkernels and generalized per-thread functions. It also includes a move of SILU and GELU operations to a dedicated unary-ops module.