IO stream double precision

The extern C shared library interface now reports in double precision, under all build settings. More importantly than double reporting precision, IO streams for unitary matrix operators typically need to be numerically unitary to at least double tolerance. This is enforced by assuming any gate is (approximately-enough) unitary, solving for a 3-paramater representation plus global phase of the operator, and reconstructing the matrix representation from the 3 parameters and global phase in double precision (before writing to an output stream).

Full Changelog: vm6502q.v9.13.1...vm6502q.v9.13.6

sha1sum results:
c5172034c08572570b5fb5cb4aea0c51828d388a libqrack-macosx_14_0_arm64.zip
86f753feb152214a26b0da3bf3a06c7c5dc92a32 libqrack-macosx_15_0_arm64.zip
7afd0502a8ddd6141a7ba7c8e74a1a17486e90f6 libqrack-manylinux2014_x86_64.zip
75c8c2fc892ed5e7ebc0c47a724eae2ed940ec6d libqrack-manylinux_2_35_x86_64.zip
830deb81392d6be9079414ea3b5ed24894e15ee2 libqrack-manylinux_2_39_x86_64.zip
8efe7fde2b103c49b1741e068b2f4a74ec6283d2 libqrack-win-amd64.zip