Focus call-graph macrobenchmarks on analyses

[liblit]

Add JMH macro-benchmarks for the call-graph workload that measure only pointer-analysis construction, not the test-harness verification the existing end-to-end benchmark includes.

The dominant call-graph benchmark (testHelloAllEntrypoints) times a whole unit test; a JFR profile showed ~30% of its wall time is spent verifying the results (graph-integrity checks, call-graph squashing and subset assertions, stat printing) rather than in the analysis itself. That verification pollutes the timed signal with JIT'ed, IO-bound, and assertion work, inflates the AB-experiment cycle, and obscures what real WALA users pay for when building call graphs.

Split the five-phase pipeline in CallGraphTest.doCallGraphs into a build step (all five call graphs, sharing one analysis cache) and the existing verification step. The verification remains unchanged and always runs in the test path, in the same phase order, so the graph-structure and determinism assertions are exactly as before.

Add two kinds of benchmarks on top:

• per-phase single-shot benchmarks (RTA, 0-CFA, 0-1-CFA, 0-Container-CFA, 0-1-Container-CFA) for cost attribution, with the scope, class hierarchy, options, and cache built once outside the timed region;

• an ensemble benchmark that runs all five phases in sequence with no verification, preserving the real pipeline's cross-phase cache and JIT behavior.

Every benchmark returns the built call graphs so the JIT cannot elide the work. No production code is touched.

Impact using my personal development machine and Zulu Java 17: the ensemble benchmark runs in ~5.8 s versus ~8.1 s for the end-to-end probe, removing the ~30% verification noise and landing almost exactly on the profile's predicted ~70% analysis portion. The per-phase methods attribute cost across algorithms (0-1-Container-CFA dominates at ~1.9 s), giving AB experiments a cheaper, lower-variance, call-graph-specific number to track optimizations.

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 50.81%. Comparing base (36ae5ff) to head (971b959).

Additional details and impacted files

@@            Coverage Diff            @@
##             master    #2053   +/-   ##
=========================================
  Coverage     50.80%   50.81%           
  Complexity    12760    12760           
=========================================
  Files          1368     1368           
  Lines         83500    83500           
  Branches      14414    14414           
=========================================
+ Hits          42426    42427    +1     
  Misses        36388    36388           
+ Partials       4686     4685    -1     

☔ 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.