Implement 3-qubit decomposition

[1tnguyen]

Similar to our existing KAK decomposition for 2-q unitary matrices, we can also implement the deterministic 3-q decomposition as a XACC Circuit class.

Ref: https://github.com/quantumlib/Cirq/blob/master/cirq-core/cirq/optimizers/three_qubit_decomposition.py
It's simple enough to have a native implementation rather than requiring Cirq dependency.

This should also benefit other high-level decomposition by optimization since we can decompose to bigger blocks.

[1tnguyen]

For reference, scipy.linalg.cossin is equivalent to this LAPACK subroutine (could be used with extern declaration, I think)
http://www.netlib.org/lapack/explore-html-3.3.0-2011-01-25/cuncsd_8f_source.html