gpu: nvidia: Updated matmul labda capture

[ShanoToni]

Description

When compiling oneDNN with C++20, the current matmul executor causes errors due to the lambda capturing this.
In C++20 implicit capture of this is an error, while in C++17 explicit capture will flag as a warning.

PR adds use of the already existing compiler workaround for the lambda capture WA_THIS_COPY_CAPTURE

Example error:

src/gpu/nvidia/cudnn_matmul_executor.hpp:89:13: error: implicit capture of 'this' with a capture default of '=' is deprecated [-Werror,-Wdeprecated-this-capture]   89 |             free_runtime_scratch(matmul_impl_->has_runtime_params(),      |             ^/export/users/mkltest/ec/oneDNN-auto-ef77e8c0-f90d-f1f3-8666-a4bf010d0e2d/sources/src/gpu/nvidia/cudnn_matmul_executor.hpp:63:33: note: add an explicit capture of 'this' to capture '*this' by reference   63 |         compat::host_task(cgh, [=](const compat::interop_handle &ih) 

Checklist

General

• [ x ] Do all unit and benchdnn tests (make test and make test_benchdnn_*) pass locally for each commit?
• [ x ] Have you formatted the code using clang-format?

Bug fixes

• [ x ] Have you included information on how to reproduce the issue (either in a github issue or in this PR)?

[mgouicem]

THanks for the fix!
Could you rename the commit message to align it to the PR title?