Notebook CI Job -- report times and fix regression

[mpharrigan]

• fixes performance regression introduced in SelectHubbard refactor #1732
• After some experimentation, run on 8 processes with maxtasksperworker=1
• Report 10 slowest notebooks

[mpharrigan]

Results from serial execution on the CI: 24 mins total

392.72s /home/runner/work/Qualtran/Qualtran/qualtran/bloqs/phase_estimation/phase_estimation_of_quantum_walk.ipynb
62.10s /home/runner/work/Qualtran/Qualtran/qualtran/bloqs/rotations/quantum_variable_rotation.ipynb
25.70s /home/runner/work/Qualtran/Qualtran/qualtran/bloqs/arithmetic/error_analysis_for_fxp_arithmetic.ipynb
22.08s /home/runner/work/Qualtran/Qualtran/qualtran/bloqs/phase_estimation/kaiser_window_state.ipynb
19.59s /home/runner/work/Qualtran/Qualtran/qualtran/Adjoint.ipynb
18.66s /home/runner/work/Qualtran/Qualtran/qualtran/bloqs/state_preparation/state_preparation_via_rotation_tutorial.ipynb
15.92s /home/runner/work/Qualtran/Qualtran/qualtran/bloqs/optimization/k_xor_sat/kikuchi_guiding_state_tutorial.ipynb
12.80s /home/runner/work/Qualtran/Qualtran/qualtran/bloqs/chemistry/resource_estimation.ipynb
 9.22s /home/runner/work/Qualtran/Qualtran/qualtran/bloqs/arithmetic/t_complexity_of_comparison_gates.ipynb
 7.81s /home/runner/work/Qualtran/Qualtran/qualtran/bloqs/multiplexers/apply_gate_to_lth_target.ipynb

[mpharrigan]

After 8370d88, 16 mins total and

50.65s /home/runner/work/Qualtran/Qualtran/qualtran/bloqs/rotations/quantum_variable_rotation.ipynb
46.49s /home/runner/work/Qualtran/Qualtran/qualtran/bloqs/phase_estimation/phase_estimation_of_quantum_walk.ipynb
21.28s /home/runner/work/Qualtran/Qualtran/qualtran/bloqs/arithmetic/error_analysis_for_fxp_arithmetic.ipynb
18.51s /home/runner/work/Qualtran/Qualtran/qualtran/bloqs/phase_estimation/kaiser_window_state.ipynb
16.40s /home/runner/work/Qualtran/Qualtran/qualtran/bloqs/state_preparation/state_preparation_via_rotation_tutorial.ipynb
14.69s /home/runner/work/Qualtran/Qualtran/qualtran/bloqs/optimization/k_xor_sat/kikuchi_guiding_state_tutorial.ipynb
13.98s /home/runner/work/Qualtran/Qualtran/qualtran/Adjoint.ipynb
11.07s /home/runner/work/Qualtran/Qualtran/qualtran/bloqs/chemistry/resource_estimation.ipynb
 8.55s /home/runner/work/Qualtran/Qualtran/qualtran/bloqs/arithmetic/t_complexity_of_comparison_gates.ipynb
 7.14s /home/runner/work/Qualtran/Qualtran/qualtran/cirq_interop/cirq_interop.ipynb

[mpharrigan]

with 8 workers, total time of the execution step is 6m17s

[mpharrigan]

without maxtasksperchild=1, the total time of the execution step is 6m22s and doesn't show any of the stdout until the very end. I'll keep that change in

[mhucka]

The way the argument parser is currently defined, if the user does not supply a value for --n-workers, the value will be None, which multiprocessing.Pool(…) will interpret as "use as many as there are CPU cores on the machine". Although it's true that the docstring above is technically correct, the reader would have to know about the behavior of multiprocessing.Pool(None) to deduce how many workers will be used. IMHO a sentence here would be helpful to explain the connection between not providing a value and the resulting number of processes used.

[mhucka]

I still suggest a note in the docstring to mention the behavior, but it's up to you.

[mpharrigan]

yep i'll add that

[mhucka]

LGTM, with a suggestion about adding a sentence to the docstring.