Sparse CSR CPU: implement addmm(dense, sparse, sparse) -> dense (#73076)
Summary:
This PR adds a possibility to multiply two sparse matrices and add the result of a product to a dense matrix.
It uses [MKL spmmd function](https://www.intel.com/content/www/us/en/develop/documentation/onemkl-developer-reference-c/top/blas-and-sparse-blas-routines/inspector-executor-sparse-blas-routines/inspector-executor-sparse-blas-execution-routines/mkl-sparse-spmmd.html) and only CPU path is implemented for now.
Ref. https://github.com/pytorch/pytorch/issues/60858
Pull Request resolved: https://github.com/pytorch/pytorch/pull/73076
Reviewed By: mikaylagawarecki
Differential Revision: D34342993
Pulled By: cpuhrsch
fbshipit-source-id: 5e5ea67cb92fbaa4d4c0eaf61e85019972989a21
(cherry picked from commit 62b8dc730e6a6736f5c03ac09eac5223cd9706cf)