Support backprop in default.mixed device for PyTorch

[eddddddy]

Context:
The third PR in a series to support backpropagation in the default.mixed device.

Description of the Change:
Now backpropagation is supported in the default.mixed device for the PyTorch interface.

Benefits:
Better integration with auto-differentiation frameworks.

Possible Drawbacks:

Related GitHub Issues:
#1528

[github-actions]

Hello. You may have forgotten to update the changelog!
Please edit doc/releases/changelog-dev.md with:

• A one-to-two sentence description of the change. You may include a small working example for new features.
• A link back to this PR.
• Your name (or GitHub username) in the contributors section.

[codecov]

Codecov Report

Merging #2680 (2b3a574) into master (553e049) will increase coverage by 0.00%.
The diff coverage is 100.00%.

@@           Coverage Diff           @@
##           master    #2680   +/-   ##
=======================================
  Coverage   99.61%   99.61%           
=======================================
  Files         251      251           
  Lines       20647    20667   +20     
=======================================
+ Hits        20567    20587   +20     
  Misses         80       80           

Impacted Files	Coverage Δ
pennylane/devices/default_mixed.py	100.00% <100.00%> (ø)
pennylane/math/single_dispatch.py	99.04% <100.00%> (+0.06%)	⬆️

---

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update 553e049...2b3a574. Read the comment docs.

[antalszava]

Looks good @eddddddy! 🎉 🥳

One main feedback is the fact that users might not able to specify Torch devices in a versatile manner, as they would with default.qubit.torch. Having said that, it may be worth merging this as is because it should work for CPU and then another PR could address adding GPU support. This should just be documented and explained well (changelog and/or device docs).

[antalszava]

[sc-21103]