Unitarity enforcement option for intermediate condition.

[dreamer2368]

• The unitarity of intermediate conditions can be enforced by specifying the input optim_unitarize_interm_ic=true.
• This currently performs modified Gram-Schmidt operation on all processes, which needs parallelization in future.

[dreamer2368]

Your code looks fine, but I don't understand why the discontinuity reported after rollout becomes non-zero. This happens, e.g., in the case results/gate_swap02/swap02.cfg when optim_unitarize_interm_ic=true. Setting that variable to false gives zero rollout discontinuities, as expected.

Added a commit to fix the rollout result.

[dreamer2368]

Three main changes are added:

• inner product convention is changed and reflected throught all unitarization routines.
• Some necessary memory clean-up routines are called in order to avoid memory leak.
• In the beginning I thought I implemented classic GS, but it turns out to be the modified one. The gradient is formulated with respect to the classic one, which costs less memory and has the equivalent gradient. In this commit, the unitarization routine reflects the conventional loop order of the modified GS.