feat(benchmarks): add model matrix run script

[ShuxinLin]

Description

Adds an executable benchmarks/run.sh launcher for running the lite scenario suite across the configured eight-model Stirrup benchmark matrix. The script accepts scenario and leaderboard roots via positional arguments or environment variables, validates required inputs, skips existing trajectories, and continues across individual scenario errors.

Type of Change

• New Benchmark Scenario (Industry/Asset type)
• Evaluation Metric / Scorer
• Agentic Orchestration Logic (ReAct, Plan-Execute, etc.)
• Infrastructure / Tooling Improvement

Industry Relevance

Provides a repeatable model-comparison entry point and consistent output layout for benchmark trajectories, reports, and Stirrup workspaces.

Related Issues

• None

Testing & Validation

• Unit Tests: 43 passed in src/benchmark/tests/test_scenario_suite_runner.py.
• Scenario Validation: Mocked uv and verified all eight correctly quoted runner invocations.
• Shell Validation: bash -n benchmarks/run.sh and missing-argument behavior passed.
• Data Integrity: No benchmark data, PII, credentials, or sensitive industrial data added.

Checklist

• The script follows repository shell conventions and passes syntax validation.
• I have performed a self-review.
• The script includes inline usage guidance for required inputs.
• I have signed off the commit (DCO).