Fix: Add DSP gemm pack with auto memcpy to buffer

[Cstandardlib]

What's changed?

• Temporary solution to make it work for nonlocal gemm on DSP.
• Add a gemm pack version, with auto memory buffer inside.

[Copilot]

Pull request overview

This PR updates the DSP GEMM path for complex types to use new “pack” variants that internally allocate DSP buffers and perform host↔DSP copies, aiming to make nonlocal GEMM work on DSP as a temporary solution.

Changes:

• Switched DSP complex GEMM calls to cgemm_pack_mth_ / zgemm_pack_mth_ in the BLAS connector.
• Added DSP connector declarations and implementations for the new packed GEMM entry points.
• Adjusted DSP connector header guard trailing #endifs.

Reviewed changes

Copilot reviewed 3 out of 3 changed files in this pull request and generated 5 comments.

File	Description
source/source_base/module_external/blas_connector_matrix.cpp	Routes DSP complex GEMM calls through new packed DSP GEMM helpers.
source/source_base/kernels/dsp/dsp_connector.h	Declares the new packed DSP GEMM APIs (and adjusts header guard endings).
source/source_base/kernels/dsp/dsp_connector.cpp	Implements packed DSP GEMM helpers with DSP-side allocations and memcpy.

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.