Add TaskDirReader for harbor datasets

[pan-x-c]

Description

This pull request adds support for reading tasksets organized as directories (folder-style tasksets), such as those used in Harbor-style benchmarks. It introduces a new TaskDirReader class, integrates it into the buffer reading system, and provides comprehensive tests for its functionality. Additionally, it generalizes the task scheduler to support any BufferReader implementation, not just file-based readers.

New folder-style taskset reader:

• Added the TaskDirReader class in trinity/buffer/reader/task_dir_reader.py, allowing Trinity to read datasets where each task is represented by a directory. The reader supports optional indexing, batch reading, and integrates with selectors.
• Registered the new "task_dir" storage type in the buffer reader registry (trinity/buffer/reader/__init__.py) and added TASK_DIR to the StorageType enum (trinity/common/constants.py). [1] [2]
• Updated the optional dependencies comment in pyproject.toml to mention the harbor package for Harbor dataset support.

Testing and validation:

• Added a comprehensive test suite for TaskDirReader in tests/buffer/task_dir_reader_test.py, covering reading, indexing, and resume functionality.

Generalization and integration:

• Generalized the TaskScheduler to accept any BufferReader implementation, not just FileReader, making it compatible with the new TaskDirReader. [1] [2]

Checklist

Please check the following items before code is ready to be reviewed.

• Code has passed all tests
• Docstrings have been added/updated in Google Style
• Documentation has been updated
• Code is ready for review