Investigate 64 bit compatibility

[ptoffy]

Right now the library uses the native representation of integers of each of the sub-libraries, so 32 bit for BLAS and LAPACK and 64 bit for Accelerate's vDSP functions. This is somewhat of a problem because we can't use BLAS/LAPACK to implement vDSP and are missing out on some performance gains due to having to implement them manually (e.g. https://github.com/brokenhandsio/accelerate-linux/blob/main/Sources/AccelerateLinux/VectorOps/VectorBasicOps.swift#L255). We should activate OPENBLAS_USE64BITINT (for BLAS) and LAPACK_ILP64 (LAPACK) to use 64 bit integers in those libraries too