feat: add parallel uncertainty fraction mapping

[ConradBitt]

Description

This PR implements the calculation of the uncertainty exponent for attraction basins, using the computed basin as a reference and a DiscreteDynamicalSystem mapping to check if the perturbation leads to a classification different from that already performed by the escape_analysis function.

Main Technical Changes

• Main Implementation: Added the uncertainty_fraction_mapping function to handle uniform sampling perturbations on disk and classify basin boundary uncertainty.
• High-Performance Parallelization: Leveraged joblib to parallelize processing at the epsilon-scale level, drastically reducing execution time.
• Thread-Safe PRNG: Integrated numpy.random.SeedSequence.spawn to guarantee strictly isolated and reproducible random streams across concurrent worker processes.
• Early Exit Optimization: Integrated a short-circuit check into the point perturbation loops to immediately flag a coordinate as uncertain as soon as the threshold mismatch condition is met, avoiding redundant path calculations.
• Documentation: Fully documented functions following the standard NumPy/SciPy style guide with internal English commentary.

Verification and Testing

• Validated performance, scalability, and array shape consistency directly through the notebook environment asset test_uncertainty_exponent.ipynb.