[NDTensors] Array storage combiner contraction refactor

[mtfishman]

With this PR, combining and uncombining of dense and block sparse tensors using the new array storage design work on a wide range of tests. I also refactored the code into two layers, one layer that takes a tensor and unwraps the indices, and another that does the combiner contraction on either Array or BlockSparseArray. I think it is a lot more readable now, and the tensor layer is very minimal.

The main limitation is that information about the QNs is not available at the BlockSparseArray level so the autofermion code had to be disabled for now. I'm sure we can find a nice solution to that, either with a callback function storing information about the QNs or with an AbstractBlockSparseArray subtype that stores QN information.

[codecov-commenter]

Codecov Report

All modified and coverable lines are covered by tests ✅

Comparison is base (063ccbe) 85.36% compared to head (b015e96) 54.56%.
Report is 2 commits behind head on main.

❗ Current head b015e96 differs from pull request most recent head da76c3d. Consider uploading reports for the commit da76c3d to get more accurate results

❗ Your organization needs to install the Codecov GitHub app to enable full functionality.

Additional details and impacted files

@@             Coverage Diff             @@
##             main    #1237       +/-   ##
===========================================
- Coverage   85.36%   54.56%   -30.81%     
===========================================
  Files          89       88        -1     
  Lines        8445     8392       -53     
===========================================
- Hits         7209     4579     -2630     
- Misses       1236     3813     +2577     

Files	Coverage Δ
src/mps/dmrg.jl	0.00% <ø> (-83.59%)	⬇️

... and 36 files with indirect coverage changes

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.