[SPARSE] Add support for sparse GEMM using MKLCPU

[Rbiessy]

Description

Add support for sparse GEMM using MKLCPU.
This relies on #402 and #374

Checklist

All Submissions

• Do all unit tests pass locally? Attach a log.
• Have you formatted the code using clang-format?

New features

• Have you provided motivation for adding a new feature?
• Have you added relevant tests?

[Rbiessy]

Log using a recent close-source MKL and DPCPP builds, all tests are passing: sparse_gemm_log.txt
I confirmed that with the 2023.2.0 release everything is building and the tests are either skipped or are passing.

[Rbiessy]

I have rebased the branch and confirmed that this is still building and running fine with oneapi 2023.2.0 as well as with recent DPCPP + MKL builds (see log above).

[noffermans]

This PR looks good to me! I mostly had questions for my own sake.
I'll wait to see whether @gajanan-choudhary wants to review the PR himself before giving approval.

[FMarno]

I think it looks good overall. Feel free to ignore my comment.

[gajanan-choudhary]

Once you add the second optimize_gemm() API, I will approve. Overall the changes look fine.
I just need to take a closer look at the gemm unit test to understand ldb/ldc/inner dim/outer dim, etc. and the combinations that play out in there, but that is not a blocker for the PR.

[gajanan-choudhary]

LGTM!