Add packaging, testing, and error handling

[geraldaddey]

This pull request enhances the project's production readiness by introducing a formal packaging structure, comprehensive test suite, and robust error handling. The project is now configured with pyproject.toml, making it an installable package with defined dependencies and caveman/caveman-nlp command-line entry points.

A full test suite using pytest has been implemented to ensure reliability, with mocked tests for API interactions. The core scripts have also been refactored to include input validation and gracefully handle potential runtime issues, such as API failures, file I/O errors, and invalid user input.

Key Changes:

1. Strict Input Validation in compress_text and decompress_text:

   • Both compress_text and decompress_text functions now include explicit checks to ensure that the
     input text is not empty or composed solely of whitespace. Invalid inputs trigger a descriptive
     error message to sys.stderr and cause the script to exit gracefully with a SystemExit status
     code of 1.

2. Improved Error Handling for File Operations in main function:

   • Permission Handling: The main function now specifically catches PermissionError during file
     read/write operations, providing clearer feedback to the user if file system permissions are
     inadequate.
   • Encoding Issues: UnicodeDecodeError is now handled when reading input files, alerting users to
     potential problems with file encoding.
   • General File Errors: A broader Exception catch-all is implemented for file operations, ensuring
     graceful failure and informative messages for unexpected file-related issues.
   • Output Directory Creation: The script now automatically creates necessary parent directories
     for output files if they don't already exist, preventing errors during output file creation.

Impact:

The compression logic is more more resilient to common user errors and environmental
factors. Users will benefit from more precise error messages, improving the script's reliability and
user experience by guiding them through troubleshooting.

[pacohope]

The beautiful irony of the long-winded, AI-authored PR comment on the caveman-compression package. It would appear we care more about being terse when LLMs are reading than when humans are.

In any human-authored PR, the phrase

Permission Handling: The main function now specifically catches PermissionError during file read/write operations, providing clearer feedback to the user if file system permissions are inadequate.

would be

Add handling for PermissionError on file operations.