[SPARSE] Add support for sparse gemv with MKLCPU

[Rbiessy]

Description

This first PR is only adding support for sparse gemv and the necessary functions to set it up. This is only adding the MKLCPU backend for now. Compliant with the provisional oneMKL specification v1.3.
The operations trsv and gemm are throwing unsupported for now.

Note that the examples are based on the oneapi dpcpp examples (like the other domains).

All Submissions

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

New interfaces

• [N/A] Have you provided motivation for adding a new feature as part of RFC and
  it was accepted? # (RFC)
• What version of oneAPI spec the interface is targeted?
  • This PR makes assumptions on the oneAPI sparse spec that are not applied yet. This matches with the close-source Sparse oneMKL spec (only some functions are not added yet).
• Complete New features checklist

New features

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

[gajanan-choudhary]

Did a partial review. Great progress here, @Rbiessy. I'll continue reviewing. Some small changes might be needed for now as a first step.

[gajanan-choudhary]

Did another round of review. It's looking good. I'll still need another round or two!

[Rbiessy]

Made some more changes to follow some spec updates and make it easier to add MKLGPU backend in the future. Please let me know if you have more feedback.

[mkrainiuk]

Thanks for the great work on new domain enabling! I have several questions and suggestions.

[Rbiessy]

@gajanan-choudhary I am sorry I squashed all the commits and forgot to keep the gemv commit separate. In a nutshell the gemv commit was only calling the MKLCPU gemv and optimize_gemv functions in src/sparse_blas/backends/mkl_common/mkl_operations.cxx and adding tests. You can still see the old diff here.

[Rbiessy]

I have confirmed the tests are still passing locally using 2023.2.0 oneapi package: tests_cpu.txt
Tests log using a MKL nightly that supports complex: tests_cpu_nightly.txt

Following our discussions I have removed the API for symv, trmv and gemvdot.

[Rbiessy]

@mmeterel I think I understand better your concern about adding API that throw unsupported now. If the risk is that users may run into unexpected errors, is adding this line in the README helping?
From my perspective it was more efficient to add all the APIs in one go and the risk of people using it is low given that sparse_blas is disabled by default.

[gajanan-choudhary]

PR is approved from my side, pending the only discussion about header naming (sparse_blas.hpp versus spblas.hpp).

[mkrainiuk]

Looks good to me, thank you!