Option to Sandbox/Specify Dataset Save Path

[kalenmike]

Added the parameter --sandbox which accepts a path, when used this overwrites the project path. When training is run datasets are downloaded to this directory and all training runs are saved in a subdirectory using the --name.

I took the opportunity to refactor some of the code to make my changes easier to read, and create unit tests to verify the datasets are downloading correctly.

🛠️ PR Summary

_{Made with ❤️ by Ultralytics Actions}

🌟 Summary

Introduction of unit tests for dataset checking and file extension validation in YOLOv5 repo.

📊 Key Changes

• 🧪 Added tests/README.md to explain how to run unit tests.
• 🚀 Created two new test files: test_check_datasets.py and test_check_suffix.py to validate dataset checking and file suffix utility functions.
• 🔧 Modified train.py to introduce --sandbox CLI argument allowing datasets and outputs to be saved in a custom directory.
• 🛠️ Updated utils/general.py with enhancements for dataset path handling and new utility functions for checking URLs and file extensions.

🎯 Purpose & Impact

• 🔍 The new tests are aimed at preventing future regressions by ensuring that dataset checking and file extension verification behave as expected.
• 📁 The --sandbox argument in train.py offers users more flexibility by specifying a custom directory for saving datasets and training outputs, improving data management.
• 💡 Enhancements to utils/general.py make code more robust in handling download URLs and checking file extensions, reducing potential errors and streamlining data preparation steps.

[kalenmike]

@glenn-jocher I submitted a simpler pull request that adds this feature with less refactoring of the code.
#4774