Two Metal crash fixes for anyone running long-context local inference on Apple silicon.

Build b10691 fixes a null-pipeline crash for F16 src1 mul_mat and mul_mat_id by declining unsupported shapes in supports_op and aborting on a nil pipeline in encoder_set_pipeline, sharing the mm dispatch between both paths so support checks and execution can't disagree. Build b10690 the same morning fixes a separate crash where a context shift on an unquantized K cache copied a Hadamard matrix to a k_rot tensor with no buffer assigned. (GitHub)