Update constraints.py

[Jonathan-Lindley]

Hello, I have been working on making various changes to the code:

Added code implementing the Q condition

Made various corrections to ensure code runs as intended (e.g. adding required if/else statements which were absent, fixing a normalisation factor)

There is an issue with dimensionality in the existing code - when running certain constraints, you input a rank-4 tensor and get out a rank-6 tensor. This then causes an issue when going to the next constraint in the sequence as the expected input is a rank-4 tensor. To fix this I have added code to convert the rank-6 tensors into rank-4 tensors, but this should be checked.

Added code to the find_closest_sdp function which reshapes rank-4 n x n x n x n tensor into matrix with dimensions n2 x n2. This should allow the function to run. The matrix is then reshaped back into a rank-4 tensor after the algorithm is applied.

[PaulWAyers]

The rank-4 to rank-6 tensor is expected. There pull-back then converts rank-6 to rank-4.

So the workflow is rank-4 to rank-6 to rank-2. Force rank-2 tensor to be positive semidefinite with the correct trace; take the corrected rank-2 tensor, conver it to rank-6, then pull back to rank-4.