Fix SCAL on x86 and RISCV_GENERIC

[martin-frbg]

fixes two oversights from #4807

[mattip]

That was quick, thanks!

[mattip]

Dunno if you want to fix it, there are some errors with Windows_cl starting here

utest\test_gemv.c(22,15): error C2057: expected constant expression

I think that is coming from declarations like this

OpenBLAS/utest/test_gemv.c

Lines 21 to 23 in fb7c53c

	float A[N * N];
	float X[N];
	float Y[N];

where N is not a constant expression.

[mattip]

Maybe you could #define N instead

[martin-frbg]

thanks for the pointer - I had seen the failure but did not find the actual error when wading through the Azure log, so was hoping it would be the same C99 thing that broke the manylinux build. indeed these should be defines (but at least that wasn't my fault except for not noticing in review)