feat: Comprehensive Analysis Runner (M121)

[hlin99]

Problem

Users must run multiple CLI subcommands (summary, tail, jitter, convergence, health-check, outlier-impact, scorecard) individually to get a complete picture of their benchmark data quality and results.

Solution

Add a comprehensive subcommand that orchestrates multiple analysis modules in a single invocation and produces a unified report with aggregated findings.

Implementation

• ComprehensiveRunner class in comprehensive.py
• ComprehensiveConfig, AnalysisSection, ComprehensiveReport Pydantic models
• Default suite: summary, SLA check, tail analysis, jitter, convergence, health check, outlier impact, scorecard
• Configurable sections via --sections / --skip flags
• Aggregated findings: warnings, recommendations, key metrics from all analyses
• Overall health summary with composite grade
• CLI comprehensive subcommand with table + JSON output
• Programmatic run_comprehensive() API

Acceptance Criteria

• comprehensive subcommand produces unified report
• Sections are independently toggleable
• JSON output includes all section results
• ~24 new tests
• All existing tests pass