Update a workaround to gemm issue in OneMKL

[antonwolfy]

The PR proposes to update the workaround implemented in #2082, based on new input from OneMKL team.
It assumes to enable the w/a only in case when an input array does not have 16 bytes alignment in the memory and to remove any check on size of input array.
Also the w/a is extended to gemm_batch functionl, since it might be impacted.

Additionally an explicit test is added to verify the w/a scenario.

• Have you provided a meaningful PR description?
• Have you added a test, reproducer or referred to issue with a reproducer?
• Have you tested your changes locally for CPU and GPU devices?
• Have you made sure that new changes do not introduce compiler warnings?
• Have you checked performance impact of proposed changes?
• If this PR is a work in progress, are you filing the PR as a draft?

[github-actions]

View rendered docs @ https://intelpython.github.io/dpnp/pull/2096/index.html

[antonwolfy]

The issue has been resolved by OneMKL team, no need for a workaround anymore.