Add native PauliRot implementation in LightningKokkos [sc-71642]

[vincentmr]

Before submitting

Please complete the following checklist when submitting a PR:

• All new features must include a unit test.
  If you've fixed a bug or added code that should be tested, add a test to the
  tests directory!

• All new functions and code must be clearly commented and documented.
  If you do make documentation changes, make sure that the docs build and
  render correctly by running make docs.

• Ensure that the test suite passes, by running make test.

• Add a new entry to the .github/CHANGELOG.md file, summarizing the
  change, and including a link back to the PR.

• Ensure that code is properly formatted by running make format.

When all the above are checked, delete everything above the dashed
line and fill in the pull request template.

---

Context:
Pauli rotations come up in many places, and importantly in the time evolution of qchem Hamiltonians. It is therefore worth considering ways to accelerate their execution.

Description of the Change:
Implement applyPauliRot. Invoke applyPauliRot directly from the SV class and add bindings to the Python layer.

Benefits:
Faster Pauli rotations. I performed a benchmark on random PauliRotations (runtime > 1.0 sec and at least 5 of them) through the Python layer. The data remains noisy with 5 samples because the performance varies depending on the specific "XYZ" sequence (which translates into more or less predictable memory access patterns). Overall, we see an advantage for 3+ qubits and up.

I performed the same benchmark on an A100 card with the Kokkos-CUDA backend, but using at least 500 samples since the absolute timings quite small and get the following speed-ups.

Using a full workflow such as

    @qml.qnode(dev, diff_method=None)
    def circuit():
        qml.TrotterProduct(ham, time=1.0, n=1, order=2)
        return qml.state()

to benchmark, we obtain timings as follows

For large enough molecules (>= 20 qubits, >= 1000 terms), the new PauliRot kernels have a clear advantage which only grows with molecular size. It is worth noting that with L-Kokkos-CUDA, even at the (24/10k) scale, evaluating the circuit is not the main bottleneck which is why it takes about the same time simulating HCN (2.64 sec. apply_lightning vs 32.5 sec. QNode) and N2N2 (7.51 sec. apply_lightning vs 36.4 sec. QNode).

Possible Drawbacks:

Related GitHub Issues:
[sc-69801]

[codecov]

Codecov Report

Attention: Patch coverage is 88.63636% with 10 lines in your changes missing coverage. Please review.

Project coverage is 97.29%. Comparing base (d5ffb0c) to head (391d0da).
Report is 72 commits behind head on master.

Files with missing lines	Patch %	Lines
...nylane_lightning/lightning_kokkos/_state_vector.py	0.00%	6 Missing ⚠️
...ane_lightning/lightning_kokkos/lightning_kokkos.py	20.00%	4 Missing ⚠️

Additional details and impacted files

@@            Coverage Diff             @@
##           master     #855      +/-   ##
==========================================
+ Coverage   96.24%   97.29%   +1.04%     
==========================================
  Files         212      168      -44     
  Lines       28109    21118    -6991     
==========================================
- Hits        27054    20547    -6507     
+ Misses       1055      571     -484     

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

[AmintorDusko]

Great job! I'm ready to approve. But first, could you please double-check the codecov coverage complains?

[AmintorDusko]

Thank you for checking!

[LuisAlfredoNu]

Nice job. Thank you @vincentmr. Only a couple of comments

[LuisAlfredoNu]

Good achievement of performance.Thank you @vincentmr 🚀