Oxidize NLocal & family

[Cryoris]

Summary

Part of #13046: create the n_local function and port the internals to Rust. This will include other parts of the NLocal family, such as TwoLocal, EfficientSU2, RealAmplitudes etc. EvolvedOperatorAnsatz is potentially a bit more involved and might be done in a follow-up.

Details and comments

The new code is particularly fast when dealing with gates we already have in Rust and that don't need any special re-parameterization. In this case:

the new code is ~22x faster for 100 qubits, full entanglement and 10 reps

num qubits: 10, speedup: 9.075
main: 0.005 +- 0.004
this: 0.001 +- 0.000

num qubits: 50, speedup: 18.138
main: 0.063 +- 0.014
this: 0.003 +- 0.003

num qubits: 100, speedup: 22.011
main: 0.251 +- 0.034
this: 0.011 +- 0.007

num qubits: 200, speedup: 26.829
main: 0.957 +- 0.034
this: 0.036 +- 0.010

But if custom gates or special parameterizations are used (e.g. RXGate(x + 3 * y)) we need to go back to Python space to correctly handle the parameter logic. This slows down the code and in this case:

the new code is ~3.5x faster for 100 qubits, linear entanglement (full took too long) and 10 reps

num qubits: 10, speedup: 4.594
main: 0.027 +- 0.006
this: 0.006 +- 0.000

num qubits: 50, speedup: 3.947
main: 0.162 +- 0.006
this: 0.041 +- 0.011

num qubits: 100, speedup: 3.666
main: 0.366 +- 0.031
this: 0.100 +- 0.019

To do

• Improve docstrings
• Make the entanglement handling rust-side a bit better to understand instead of a 4-times nested vec
• Use n_local in NLocal (i.e. fast path for Python code) This might not be possible since NLocal works with QuantumCircuits as blocks, but n_local needs gates. Functionally it'd be the same but the circuits are nested differently, which is not backward compatible. We could add a fast-path in case the inputs are not circuits but gates.
• Add the family members, like efficient_su2 etc.

[coveralls]

Pull Request Test Coverage Report for Build 11385891601

Warning: This coverage report may be inaccurate.

This pull request's base commit is no longer the HEAD commit of its target branch. This means it includes changes from outside the original pull request, including, potentially, unrelated coverage changes.

• For more information on this, see Tracking coverage changes with pull request builds.
• To avoid this issue with future PRs, see these Recommended CI Configurations.
• For a quick fix, rebase this PR at GitHub. Your next report should be accurate.

Details

• 528 of 540 (97.78%) changed or added relevant lines in 14 files are covered.
• 48 unchanged lines in 13 files lost coverage.
• Overall coverage increased (+0.06%) to 88.713%

---

Changes Missing Coverage	Covered Lines	Changed/Added Lines	%
qiskit/circuit/library/n_local/excitation_preserving.py	15	16	93.75%
crates/accelerate/src/circuit_library/n_local/blocks.rs	92	95	96.84%
qiskit/circuit/library/n_local/n_local.py	87	95	91.58%

Files with Coverage Reduction	New Missed Lines	%
crates/accelerate/src/two_qubit_decompose.rs	1	91.54%
qiskit/circuit/library/standard_gates/u.py	1	93.0%
qiskit/circuit/library/standard_gates/rz.py	2	95.83%
qiskit/synthesis/stabilizer/stabilizer_decompose.py	2	96.77%
qiskit/circuit/library/standard_gates/ry.py	2	95.95%
qiskit/circuit/library/standard_gates/u3.py	2	97.06%
qiskit/synthesis/unitary/qsd.py	2	97.56%
qiskit/circuit/library/standard_gates/rx.py	2	96.05%
qiskit/transpiler/passes/scheduling/base_scheduler.py	5	88.37%
crates/qasm2/src/parse.rs	6	97.61%

Totals
Change from base Build 11372663133:	0.06%
Covered Lines:	73694
Relevant Lines:	83070

---

💛 - Coveralls

[qiskit-bot]

One or more of the following people are relevant to this code:

• @Cryoris
• @Qiskit/terra-core
• @ajavadia