Add comprehensive code review documentation for TrackLab

[veillette]

Summary

This PR adds a detailed code review document (CODE_REVIEW.md) that provides a comprehensive analysis of the TrackLab codebase across architecture, code quality, correctness, performance, security, and maintainability dimensions.

Changes

• Added CODE_REVIEW.md: A 215-line code review document covering:
  • Architecture & Design: Evaluates model-view separation, immutable data flow, configuration management, and identifies design concerns around view-layer validation and unused feature stubs
  • Code Quality: Assesses naming conventions, TypeScript usage, and identifies issues with verbose kinematics computation, inconsistent property linking patterns, and magic numbers
  • Bugs & Correctness: Documents 7 issues ranging from critical (unvalidated OpenCV ROI dimensions that can crash WASM) to minor (floating-point rounding inconsistencies)
  • Performance: Identifies 4 optimization opportunities including full kinematics recomputation on track changes and O(n) array operations in the auto-tracker trail
  • Security: Flags the critical ROI validation issue, silent error handling in cross-origin video loading, and untyped WASM module access
  • Maintainability: Highlights missing dispose() methods in 3 view components, a timer leak in WebcamPanel, and lack of i18n key validation

Key Findings

Critical Issues (3):

• OpenCV ROI dimensions not validated for negative/zero values
• Duplicate point rejection silently discards user data without feedback
• Missing dispose() methods on view components creates inconsistent cleanup patterns

Moderate Issues (7):

• View-layer validation logic that belongs in the model
• Verbose, duplicated kinematics computation code
• Full kinematics recomputation triggered by any track change
• Auto-tracker race condition when tracks are deleted during async initialization
• Inconsistent X/Y axis drag direction behavior

Minor Issues (6):

• Floating-point frame rounding inconsistencies
• O(n) Array.shift() operations in auto-tracker trail
• Magic numbers in constants without clear derivation
• Silent error handling for cross-origin video failures

Notes

This is a documentation-only change that provides actionable feedback for future improvements. No source code modifications are included in this PR.

https://claude.ai/code/session_01V2288fov5zhNrnkjT8uACW