linalg: enable 80-bit extended precision for whole library, `xdp`

[perazz]

Enable xdp precision for BLAS, LAPACK and whole linear algebra library. This PR now seamlessly enables both 80-bit extended-precision (xdp) and 128-bit quadruple precision (qp) versions of the liear algebra library to be built.

• [x] extended precision functions templated for whole blas, lapack
• [x] new initials defined:

real entities -> s, d, x, q complex entities -> c, z, y, w

• [x] whole functionality including BLAS/LAPACK interfaces, as well as the whole testsuite, have now removed the constraints to not use xdp.
• [x] both "high" precision LAPACK modules share use of 64-bit double precision for the algorithms that require a "lower precision" kind

cc: @jvdp1 @jalvesz @fortran-lang/stdlib