Add malicious security support to oneshot bench

[andyleiserson]

We no longer need to (try to) exercise the complete step space in collect_steps, so this also reduces the set of supported configs to (semi-honest, malicious) × (32 or 256 breakdowns), all with 3-bit trigger value and 3-bit saturating sum. (Timestamp and output value also remain fixed at BA20 and BA32, respectively.)

[andyleiserson]

@tyurek FYI since we were talking about profiling today: I use this for flamegraphs with cargo flamegraph --root --bench oneshot_ipa --features "enable-benches disable-metrics multi-threading" -- -n 100000 -j 8. Be aware, however, that the bottlenecks indicated by a CPU profile with in-memory networking can be quite different than the bottlenecks in a real cluster.

[codecov]

Codecov Report

Attention: Patch coverage is 29.68750% with 45 lines in your changes missing coverage. Please review.

Project coverage is 93.46%. Comparing base (52c58eb) to head (098a663).

Files with missing lines	Patch %	Lines
ipa-core/src/test_fixture/ipa.rs	29.68%	45 Missing ⚠️

Additional details and impacted files

@@            Coverage Diff             @@
##             main    #1405      +/-   ##
==========================================
- Coverage   93.49%   93.46%   -0.04%     
==========================================
  Files         223      223              
  Lines       38396    38434      +38     
==========================================
+ Hits        35900    35921      +21     
- Misses       2496     2513      +17     

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