This changes the kernels to pack full SVE vectors and reduces the
overall complexity of the inner GEMM loop.

* Update to 20.1.8

* fix PATH to avoid the obsolete LLVM19 that appeared in the preinstalled msvc folder hierarchy

Fix compilation of the NeoverseN2 SBGEMM kernel

Multi-thread Performance Improvement of GEMM on NeoverseV1 with DIVIDE_RATE=1

Adjust multithreading threshold in S/DGER and add an intermediate step

Work around gcc15.1 on POWER misoptimizing DGEMV at -O3

Fix bad vector zero initializer and other compiler warnings for RISC-V.

Fix building of bgemm tests on GEMM3M-capable (x86) targets

ARM64: Enable bfloat16 kernels by default

Introduce a (crude) threshold to multithreading in STRMV/DTRMV

Add NEOVERSEV2 target support

Include float-bfloat conversion functions in ONLY_CBLAS builds as well

Add optimized BGEMM for NEOVERSEN2 target

Use OpenBLAS_ROOT_DIR in OpenBLASConfig.cmake generation only if set

Update ?GEMM-to-?GEMV forwarding settings for CMake

Did a quick run around to make `TARGET=NEVOERSEV2` build successfully.

Fixes #5385

This re-uses the existing NEOVERSEN2 8x4 `sbgemm` kernel to implement `bgemm`.

Fix cmake building with cblas_bgemm

Optimized BGEMV for NEOVERSEV1 target

Fix libffi6 download in the Loongarch64_clang CI job (for now)

- Adds bgemv T based off of sbgemv T kernel
- Adds bgemv N which is slightly alterated to not use Y as an
accumulator due to the output being bf16 which results in loss of
precision
- Enables BGEMM_GEMV_FORWARD to proxy BGEMM to BGEMV with new kernels

Declare the "small" complex DOT and AXPY kernels for RISCV-ZVL256B static in addition to inline

Add cross-compilation parameters for RISCV64 targets in CMake

Fixes for some gcc warnings