feat: Protocol for DataLoader Added

[AndreyIsakov01]

feat: Add DatasetLoader protocol and RealDatasetLoader implementation

Summary

• Introduce a DatasetLoader protocol that defines the load(path, timeseries_preprocessor) interface for loading datasets from disk.
• Add RealDatasetLoader as a concrete implementation that discovers timeseries*.csv files, builds segment info from the SEGMENT_COLUMN, and constructs PandasLabeledDataProvider instances with proper annotations.
• Simplify Dataset by removing timeseries_preprocessor storage and the stub load_from_dir method -- loading responsibility is now delegated to DatasetLoader.

Changes

• pysatl_cpd/core/data_providers/dataset.py

  • Added DatasetLoader protocol with a load method signature.
  • Added RealDatasetLoader with three methods: prepare_segment_info (derives segment boundaries from SEGMENT_COLUMN), prepare_annotation (creates Annotation from file path), and load (orchestrates CSV discovery, preprocessing, and provider construction).
  • Simplified Dataset.__init__ -- removed timeseries_preprocessor parameter, removed load_from_dir stub, removed _identity helper.
  • Renamed SEGMENT_COLUMN value from "segments" to "segment" for consistency with SEGMENT_ID_COLUMN.
  • Replaced name-mangled __ attributes with conventional _ private attributes throughout PandasLabeledDataProvider and Dataset.

• pysatl_cpd/core/data_providers/__init__.py

  • Exported RealDatasetLoader from the module.