Ensure simulator runs circuit compatible with a provided noise model's properties

[Acciaccatura]

When provided a NoiseModelFromNoiseProperties noise model to simulate, ensure the circuit uses qubits and gates that are compatible with the noise properties. At a slightly lower level:

• NoiseModelFromNoiseProperties exposes its underlying NoiseProperties as a public property.
• SimulatorBase will check to make sure all the gates and qubits used by the circuit are permitted under any NoiseProperties from the associated NoiseModelFromNoiseProperties.
  • In the current state, only SuperconductingQubitsNoiseProperties contains expected_qubits or expected_gates, so the code is currently hardcoded to this class. I'm uncertain if Cirq has any plans on genericizing this behavior to NoiseProperties.

fixes #6608

[Acciaccatura]

question - this change seems to be blocked on a Docstring test failure, specifically this one:

Cirq/cirq-google/cirq_google/engine/calibration_to_noise_properties.py

Lines 21 to 27 in 780fe1f

	>>> cal = cirq_google.engine.load_median_device_calibration("rainbow")
	>>> noise_props = cirq_google.engine.noise_properties_from_calibration(
	... cal, gate_times_ns="legacy")
	>>> noise_model = cirq_google.NoiseModelFromGoogleNoiseProperties(noise_props)
	>>> simulator = cirq.Simulator(noise=noise_model)
	>>> circuit = cirq.Circuit(cirq.X(cirq.GridQubit(5, 2)))
	>>> result = simulator.simulate(circuit)

however when checking the expected gateset of the noise properties from this processor myself, it seems its failure is working as intended according to my implementation:

>>> import cirq_google
>>> cal = cirq_google.engine.load_median_device_calibration("rainbow")
>>> noise_props = cirq_google.engine.noise_properties_from_calibration(cal, gate_times_ns="legacy\
")
>>> print(noise_props.expected_gates())
{<class 'cirq.ops.measurement_gate.MeasurementGate'>, <class 'cirq.ops.swap_gates.ISwapPowGate'>, <class 'cirq.ops.fsim_gate.FSimGate'>, <class 'cirq.ops.common_gates.ZPowGate'>, <class 'cirq.ops.fsim_gate.PhasedFSimGate'>, <class 'cirq.ops.common_gates.CZPowGate'>, <class 'cirq.ops.phased_x_z_gate.PhasedXZGate'>, <class 'cirq_google.ops.sycamore_gate.SycamoreGate'>, <class 'cirq.ops.common_channels.ResetChannel'>}

(i.e. my implementation assumes cirq.X or any of its superclass gates must appear in the noise property's expected gateset, which it doesn't). is this expected, and if not, how should I change my implementation?

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 99.59%. Comparing base (4adcfb9) to head (7d3a59d).

Additional details and impacted files

@@           Coverage Diff           @@
##             main    #8139   +/-   ##
=======================================
  Coverage   99.59%   99.59%           
=======================================
  Files        1131     1131           
  Lines      103601   103642   +41     
=======================================
+ Hits       103180   103221   +41     
  Misses        421      421           

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

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.
• 📦 JS Bundle Analysis: Save yourself from yourself by tracking and limiting bundle sizes in JS merges.

[Acciaccatura]

the TestNoiseProperties created for this test might not make sense from a QC point of view just because of my limited domain knowledge - feel free to re-implement to something more sensible if this is a concern!

[mhucka]

@Acciaccatura Thank you for this effort. If you are still working on this, would you mind changing it to a draft PR?

[review-notebook-app]

Check out this pull request on 

See visual diffs & provide feedback on Jupyter Notebooks.

---

Powered by ReviewNB

[Acciaccatura]

hey @mhucka ! sorry I missed your message. this PR is ready for review - I just rebased :). i've left some comments on some areas I think might require some clarification.

[Acciaccatura]

noise models define their expected gateset explicitly - for example:

Cirq/cirq-google/cirq_google/devices/google_noise_properties.py

Lines 189 to 191 in 4adcfb9

	@classmethod
	def single_qubit_gates(cls) -> set[type]:
	return {cirq.ZPowGate, cirq.PhasedXZGate, cirq.MeasurementGate, cirq.ResetChannel}

there are some gates that are mathematically equivalent to other gates (e.g. I think PhasedXZGate(axis_phase_exponent=0, x_exponent=1, z_exponent=0) and X), but I believe this check will not allow X. is this ok, and if not is there a better alternative for testing gate equality?