feat: Step 17 — repetitions and confidence intervals for benchmarks

[PSchmitz-Valckenberg]

Adds statistical rigor to the benchmark pipeline. Each strategy can now be run N times; the report exposes mean and stddev per metric so noise-floor effects (e.g. 10% attack-success delta on N=1) are distinguishable from real differences.

Schema (V6):

• benchmarks.repetitions INTEGER NOT NULL DEFAULT 1
• benchmark_runs PK changed from (benchmark_id, strategy_type) to (benchmark_id, strategy_type, repetition_index) — allows N rows per strategy per benchmark

API:

• BenchmarkCreateRequest: optional repetitions field (1–10, default 1)
• BenchmarkReportResponse: includes repetitions count
• RunComparisonEntry: runId → runIds list, plus new aggregated field
• AggregatedStrategyMetrics: mean + stddev (null when N=1) for attackSuccessRate, falsePositiveRate, refusalRate, avgLatencyMs

Service:

• BenchmarkService.executeBenchmark: inner loop runs each strategy N times, recording a BenchmarkRun per repetition
• BenchmarkService.getReport: groups runs by strategy, aggregates using population stddev; delta vs baseline uses first NONE run as representative (unchanged semantics for N=1)
• mean/stddev helpers are package-visible statics for unit testing

Script:

• --repetitions N flag (default 3); summary table shows mean + stddev columns; N=1 stddev shown as "n/a"

Tests:

• 8 new unit tests covering mean, stddev edge cases, N=1 null stddev, N=3 aggregation with known values

[Copilot]

Pull request overview

Adds repetitions and aggregated statistics (mean/stddev) to the benchmark pipeline so benchmark comparisons can distinguish noise from real strategy deltas.

Changes:

• Introduces benchmarks.repetitions and updates benchmark_runs PK to include repetition_index (V6 migration).
• Updates benchmark execution/reporting to run each strategy N times and return per-strategy aggregated metrics in the report API.
• Updates CLI script output formatting and adds unit tests for mean/stddev/aggregation helpers.

Reviewed changes

Copilot reviewed 11 out of 11 changed files in this pull request and generated 4 comments.

Show a summary per file

File	Description
src/test/java/com/sentinelcore/service/BenchmarkStatisticsTest.java	Adds unit tests for mean/stddev and aggregation behavior (including N=1 null stddev).
src/main/resources/db/migration/V6__add_repetitions.sql	Adds benchmarks.repetitions and extends benchmark_runs PK with repetition_index.
src/main/java/com/sentinelcore/service/BenchmarkService.java	Executes each strategy multiple times; groups runs by strategy; aggregates metrics; exposes mean/stddev helpers.
src/main/java/com/sentinelcore/dto/RunComparisonEntry.java	Changes runId to runIds and adds aggregated metrics payload.
src/main/java/com/sentinelcore/dto/BenchmarkReportResponse.java	Includes repetitions in the report response.
src/main/java/com/sentinelcore/dto/BenchmarkCreateRequest.java	Adds optional validated repetitions field and a defaulting helper.
src/main/java/com/sentinelcore/dto/AggregatedStrategyMetrics.java	New DTO carrying mean/stddev per metric (stddev nullable for N=1).
src/main/java/com/sentinelcore/domain/entity/BenchmarkRun.java	Adds repetitionIndex to persist repetition identity per run.
src/main/java/com/sentinelcore/domain/entity/Benchmark.java	Adds persisted repetitions field (default 1).
src/main/java/com/sentinelcore/controller/BenchmarkController.java	Passes repetitions through to BenchmarkService.createBenchmark.
scripts/run_benchmark.sh	Adds --repetitions flag and updates summary output to use aggregated fields.

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.