Speed up test_ax_sweep_impl_minimal by mocking Sobol sensitivity MC samples

[saitcakmak]

Summary:
automl.internal.ax.ax_sweep.tests.test_ax_sweep_feature_matrix.TestAxSweepFeatureMatrix.test_ax_sweep_impl_minimal (T562950220699357) has been timing out on stress runs against TPX's 600s limit. Profiling shows the runtime is dominated by variance-based (Sobol) sensitivity analysis, not by the number of analysis cards:

• AxSweepOrchestrator.report_results is called 4x and accounts for 110.2s of a 127.2s local run (87%): _compute_and_save_analysis_cards 67.3s, legacy get_standard_plots 38.5s.
• Exclusive per-Analysis accounting: SensitivityAnalysisPlot 65.9s over 2 calls; every other analysis is under 0.5s.
• ax_parameter_sens defaults to num_mc_samples = 10**4, so each call is O(num_mc_samples * num_params) posterior-mean evaluations, multiplied by the MCMC sample count on this test's fully Bayesian (SAASBO) model. mock_botorch_optimize does not help, since it cheapens fitting, not posterior evaluation.
• It runs 4x per test: twice via get_standard_plots -> ax_parameter_sens(order="total") and twice via OverviewAnalysis -> InsightsAnalysis -> TopSurfacesAnalysis -> SensitivityAnalysisPlot, where _choose_sensitivity_order(3) selects the most expensive "second" order.

This adds an opt-in mock_sensitivity_analysis decorator / context manager to ax/utils/testing/mock.py that forces num_mc_samples=32 in compute_sobol_indices_from_model_list and compute_derivatives_from_model_list, and applies it to execute_ax_sweep_and_make_assertions. Nothing in that test asserts on sensitivity values -- only that the expected number of plots is produced -- so the reduced sample count is safe here.

Note that Ax's own TestCase.MAX_TEST_SECONDS = 120 guard does fire on this test, but the resulting TimeoutError is swallowed by Analysis.compute_result's broad except Exception (surfacing only as a "Failed to compute SensitivityAnalysisPlot" error card) and by report_results' broad handlers, so the test still reported ok. That is worth fixing separately.

Differential Revision: D118535425

[meta-codesync]

@saitcakmak has exported this pull request. If you are a Meta employee, you can view the originating Diff in D118535425.

[codecov-commenter]

Codecov Report

❌ Patch coverage is 23.52941% with 13 lines in your changes missing coverage. Please review.
✅ Project coverage is 96.57%. Comparing base (d034ba2) to head (4fe7590).

Files with missing lines	Patch %	Lines
ax/utils/testing/mock.py	23.52%	13 Missing ⚠️

Additional details and impacted files

@@            Coverage Diff             @@
##             main    #5282      +/-   ##
==========================================
- Coverage   96.59%   96.57%   -0.03%     
==========================================
  Files         621      621              
  Lines       70894    70911      +17     
==========================================
- Hits        68481    68480       -1     
- Misses       2413     2431      +18     

☔ 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.

[meta-codesync]

This pull request has been merged in 778e22f.