Implement DYNAMIC_ARCH support for riscv64.  Three cpu types are
supported, riscv64_generic, riscv64_zvl256b, riscv64_zvl128b.
The two non-generic kernels require CPU support for RVV 1.0 to
function correctly.  Detecting that a riscv64 device supports
RVV 1.0 is a little complicated as there are some boards on the
market that advertise support for V via hwcap but only support
RVV 0.7.1, which is not binary compatible with RVV 1.0.  The
approach taken is to first try hwprobe.  If hwprobe is not
available, we fall back to hwcap + an additional check to distinguish
between RVV 1.0 and RVV 0.7.1.

Tested on a VM with VLEN=256, a CanMV K230 with VLEN=128 (with only
the big core enabled), a Lichee Pi with RVV 0.7.1 and a VF2 with no
vector.

A compiler with RVV 1.0 support must be used to build OpenBLAS for
riscv64 when DYNAMIC_ARCH=1.

Signed-off-by: Mark Ryan <markdryan@rivosinc.com>

Fix NAN handling in ARM/generic SCAL; have AzureCI Windows show errors on failure

Correctly detect ARM Neoverse V2 CPUs.

* avoid setting nthreads higher than available

Disable GEMM3M for generic targets (not implemented)

POWER: Fix issues in zscal to address lapack failures

Commit def4996 fixed issues with inf and nan values in zscal,
but used FLT_MAX, where DBL_MAX or isinf() is more appropriate,
as FLT_MAX is for single precision only.
Using FLT_MAX caused test case failures in the LAPACK tests.

isinf() is consistent with the later fix 969601a1

This patch fixes following lapack failures with clang compiler on POWER.
zed.out: ZVX:   18 out of  5190 tests failed to pass the threshold
zgd.out: ZGV drivers:     25 out of   1092 tests failed to pass the threshold
zgd.out: ZGV drivers:      6 out of   1092 tests failed to pass the threshold

Fix regression SAXPY when compiler with OpenXL compiler.

SAXPY built with OpenXL regresses when compared to SAXPY
built with gcc. OpenXL compiler doesn't know that the
SAXPY inner kernel assembly is a 64 element loop and
to it the remainder loop is the main loop. It vectorizes
and interleaves the remainder to be a 48 elements per
iteration loop. With a max of 63 iterations, a 48 element
loop is mostly not going to get executed, so the 1 element
scalar loop that is the remainder after that is probably
mostly what gets executed.

This can be fixed by adding a pragma, loop interleave_count(2)
which will result in 8 element loop.

Signed-off-by: Amrita H S <amritahs@linux.vnet.ibm.com>

Loongarch64: Improving the Performance and Stability of dgemm