reengineering: benchmarking system

[iraedeus]

This pull request introduces a new benchmark_executor.py module to the pysatl_cpd/benchmark package, providing a robust framework for executing and caching benchmarks for online change-point detection (CPD) algorithms. The module enables running multiple algorithms across datasets and thresholds, with optional disk-based caching to avoid redundant computations.

Key additions and features:

Benchmarking orchestration and caching:

• Introduces the BenchmarkExecutor generic class, which coordinates the execution of benchmarks for multiple algorithms, datasets, and thresholds, supporting optional result caching via disk dumping.
• Implements a CSV-based registry and Pickle-based trace serialization to cache and retrieve benchmark results, minimizing redundant computations on subsequent runs.

Benchmark metadata management:

• Adds the BenchmarkRecord dataclass to uniquely identify each benchmark run and store metadata, including algorithm, configuration, dataset, threshold, and cache file path.

Integration with existing components:

• Integrates with OnlineAlgorithm, DataProvider, OnlineCpdSolver, and OnlineDetectionTrace from the core modules, ensuring compatibility and reusability within the CPD framework.

[codacy-production]

Not up to standards ⛔

_{TIP This summary will be updated as you push new changes. Give us feedback}