Add native tools for flattening and unflattening pytrees

[albi3ro]

Context:

Pytrees are nested data structures. Pytree tools can make it easier to handle said data structure.

Jax, optree, and other packages provide tools for handling pytrees already. To take advantage of those tools in core pennylane, we would need to make one of those packages a dependency. Instead of adding an extra dependency, we can have our own stripped-down version of pytree tools.

Description of the Change:

This PR adds tree_flatten, tree_unflatten, Structure, and Leaf to the pytrees module. It also updates qml.pytrees.register_pytree to register the type with pennylane's pytree setup as well.

>>> op = qml.adjoint(qml.Rot(1.2, 2.3, 3.4, wires=0))
>>> data, structure = flatten(op)
>>> data
[1.2, 2.3, 3.4]
>>> structure
<Tree(AdjointOperation, (), (<Tree(Rot, (<Wires = [0]>, ()), (Leaf, Leaf, Leaf))>,))>
>>> unflatten([-2, -3, -4], structure)
Adjoint(Rot(-2, -3, -4, wires=[0]))

Benefits:

Use in datasets module to serialize and de-serialize pennylane objects.

Potential future use in bind_new_parameters.

Easy extraction and reset of parameters.

Possible Drawbacks:

Not as performant as the C++ bound versions used by jax and optree.

Related GitHub Issues:

[sc-46349]

[codecov]

Codecov Report

All modified and coverable lines are covered by tests ✅

Project coverage is 99.67%. Comparing base (58668d7) to head (6fa9778).
Report is 1 commits behind head on master.

Additional details and impacted files

@@            Coverage Diff             @@
##           master    #5701      +/-   ##
==========================================
- Coverage   99.68%   99.67%   -0.01%     
==========================================
  Files         416      416              
  Lines       39049    38814     -235     
==========================================
- Hits        38925    38689     -236     
- Misses        124      125       +1     

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

[brownj85]

Overall looks great! this will be much easier to integrate than the Jax library

[mudit2812]

Looks great! Couple of suggestions.