remove non-vectorized SGEMV transpose reduce path for POWER8, restoring optimizations frpm PR4880

Fix and reinstate the Cooper Lake/Sapphire Rapids microkernel for non-transpose SBGEMV

Explicit unaligned vector load/stores in PPC64LE GEMV kernels

Temporarily disable the Cooper Lake/Sapphire Rapids microkernel for non-transpose SBGEMV

Improve performance for SGEMVN on NEONVERSEN1

Further performance improvements to non-transposed [SD]GEMV kernels for A64FX and Neoverse V1.

Use SVE kernel for S/DGEMVT for SVE machines

Optimizing the Implementation of GEMV on the RISC-V V Extension

CMake: Pass `OpenMP` compiler and linker flags through CMake targets

Specialized some scenarios, performed loop unrolling, and reduced the
number of multiplications.

spelling error where zsum_kernel is used and it should be zasum_kernel. Will not compile without fix.

SME based SGEMMDIRECT kernel uses the vector registers (z) and adding
    clobber list informs compiler not to optimize these registers.

On z-series, the current version of the GNU toolchain produces warnings
such as:
```
/usr/lib64/gcc/[...]/s390x-suse-linux/bin/ld: warning: ztrmm_kernel_RC_Z14.o: missing .note.GNU-stack section implies
executable stack
/usr/lib64/[...]/s390x-suse-linux/bin/ld: NOTE: This behaviour is deprecated and will be removed in a future version of the linker
```
To prevent this message and make sure we are future proof, add
```
	.section        .note.GNU-stack,"",@progbits
```
Also add the `.size` bit to give the asm defined functions a proper size
in the symbol table.

Signed-off-by: Egbert Eich <eich@suse.com>

Using `OpenMP::OpenMP_LANG` targets for CMake is less error-prone than
passing the compiler and linker flags manually. Furthermore, it allows
the user to customize those flags by setting `OpenMP_LANG_FLAGS`,
`OpenMP_LANG_LIB_NAMES`, and `OpenMP_omp_LIBRARY`.

Optimize aarch64 sgemm_ncopy

replace customize bf16_to_fp32 with arm neon vcvtah_f32_bf16

Optimize gemv_n_sve kernel

Add dispatch of SBGEMVNKERNEL for NEOVERSEN2 and NEOVERSEV2

This improves performance for sbgemv_t by up to 100x on NEOVERSEV1.
The geometric mean speedup is ~61x for M=N=[2,512].

Support for SGEMM_DIRECT Kernel based on SME1

* Added ARMV9SME target
* Added SGEMM_DIRECT kernel based on SME1

LoongArch64: Fixed lapack test for LA264

When the data type is single-precision real or single-precision complex,
converting it to double precision does not prevent overflow (as exposed in LAPACK tests).
The only solution is to follow C's approach: find the maximum value in the
array and divide each element by that maximum to avoid this issue

Fixed the error when the stride is zero