Fix gauge fixing in :fixed mode for non-uniform unit cells from full SVD

[pbrehmer]

I noticed that when using :fixed mode differentiation the gauge fixing routine would error for non-uniform unit cells when using a full SVD (as opposed a IterSVD). This came down the fact that the gauge signs on the full space were initialized with the wrong spaces. Now we explicitly initialize the correct spaces which is also tested in the test/ctmrg/unitcell.jl tests.

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.

Files with missing lines	Coverage Δ
...rithms/optimization/fixed_point_differentiation.jl	94.48% <100.00%> (+0.49%)	⬆️

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.

[pbrehmer]

The random cartesian spaces test should be fixed, so hopefully now everything runs through.

I also added a line to the _rrule(::GradMode{:fixed}, ...) which ensure fixed space truncation, as suggested in #157. This should now hopefully prevent any errors related to space incompatibility during gauge fixing.

[lkdvos]

Suggested a slight reorganization of the code, otherwise happy to merge once tests pass (also happy to merge like this)