Set GEMM_PREFERED_SIZE parameter for Neoverse V1

Remove -openmp flag from XLF (since it doesn't support it).

loongarch: Optimizing the performance of the GEMM on servers

Get the l2 cache size via environment variable on confidential VM

Don't pass `-exhaustive-register-search` directly to clang compiler

Retain the bf16 in fallback versions of the NeoverseN2 -march flag

Do not run the CBLAS_?GEMM3M tests when cross-compiling with gmake

Make the new ZSCAL utest not require CBLAS

`-exhaustive-register-search` is an LLVM code generation flag that shouldn't be passed directly to clang compiler.

The CPUID(leaf:2 or leaf:0x80000006) is not supported on some confidential
VMs. As a result the get_l2_size() returns the default 512M which brings
performance issues.

Introduce the environment variable OPENBLAS_L2_SIZE provided by the user
to get the l2 cache size.

Suggested-by: "Keshavamurthy, Anil S" <anil.s.keshavamurthy@intel.com>
Signed-off-by: Chen Yu <yu.c.chen@intel.com>

Fold wiki contents into formal documentation, build-able with `mkdocs`

merge back from release branch to copy tag

Merge develop branch for 0.3.27

Update Changelog.txt for 0.3.27

Adjust SWITCH_RATIO for ZEN and apply GEMM_PREFERRED_SIZE

Clean up misplaced LAPACK files from PR4043 (in-code documentation changes only)

Add GEMM3M tests and logs to .gitignore

Fix typedef of logical in the f2c-converted LAPACK C files to match INTERFACE64