Fixed svd_flip() when used on GPU w/ PyTorch backend

[AtomicCactus]

Fixes a problem where one of the tensors is being initialized on CPU and concatenated with a GPU tensor. See #503

[codecov[bot]]

Codecov Report

Merging #504 (649c01c) into main (e45c97e) will decrease coverage by 0.03%. The diff coverage is 0.00%.

@@            Coverage Diff             @@
##             main     #504      +/-   ##
==========================================
- Coverage   86.65%   86.62%   -0.03%     
==========================================
  Files         120      120              
  Lines        7426     7426              
==========================================
- Hits         6435     6433       -2     
- Misses        991      993       +2     

Impacted Files	Coverage Δ
tensorly/tenalg/svd.py	84.12% <0.00%> (ø)

... and 1 file with indirect coverage changes

:mega: We’re building smart automated test selection to slash your CI/CD build times. Learn more

[aarmey]

Thanks for catching this, @AtomicCactus. Merging.