loschmidt.tilted_square_lattice Benchmark

[mpharrigan]

Adds the executable generation functionality for the following benchmark:

AlgorithmicBenchmark(
    domain='recirq.otoc',
    name='loschmidt.tilted_square_lattice',
    executable_family='recirq.otoc.loschmidt.tilted_square_lattice',
    spec_class=TiltedSquareLatticeLoschmidtSpec,
    executable_generator_func=get_all_tilted_square_lattice_executables,
)

The loschmidt echo runs a random unitary forward and backwards and measures how often we return to the starting state. This benchmark checks random unitaries generated on the TiltedSquareLattice topology.

Notes

This PR does not include any BenchmarkConfig's, namely scripts for launching execution and analysis.

I've also removed the data_class field from the AlgorithmicBenchmark card catalog. (1) I don't want to make this PR too big by including it (2) I want to work some of the kinks out of the design of the data analysis part of the flow without holding up committing the infrastructure for taking data.

Cirq requirements

Following #237 , this will only test the new functionality against the pre-release version of cirq where the cirq_google.workflow utilities are available.

I've noticed as well that Cirq 0.12 does not include cirq.NamedTopology. This is a case where keeping compatibility with the prior version of Cirq is not worth it, esp. since v0.13 has been out for a while. As part of this PR I've disabled the "pre" check with a note. We'll re-enable when Cirq 0.14 comes out.

[dstrain115]

I feel like I missed a change somewhere, since this is likely not the first time this snippet has appeared.
However, what is this? I am confused why we need this. Maybe a code comment would be good to explain what this is for.

[mpharrigan]

this is how we register modules for json deserialization; done after the separation of cirq_google and friends. You can see we add it to cirq's DEFAULT_RESOLVERS global variable

[mpharrigan]

added a docstring

[dstrain115]

Why do we need this monkeypatch? Is it just for test performance?
If so, how slow is the test without this?

[mpharrigan]

It's for speed and so I can assert something, namely that it was called the correct number of times.

Creating random circuits is expensive -- or at least speeding that up is outside the scope for this work. I think I wanted a test that relied on the default arguments for this function. I can add another test that dials back the number of executables and makes other assertions.

[dstrain115]

Maybe add a comment to that effect above and explain what the test is actually testing.

[mpharrigan]

added

[mpharrigan]

@dstrain115 I think I addressed your comments. PTAL if you have time

[dstrain115]

Minor comments then I approve.

[dstrain115]

Ha, it's 2022 now! Gotcha!
(optional since you started it in 2021)

[mpharrigan]

oh no! now people can skirt our onerous apache2 license because the copyright is invalid!

[dstrain115]

Maybe add a comment to that effect above and explain what the test is actually testing.

[mpharrigan]

made the requested changes!

[dstrain115]

LGTM