Fix issues #224 and #225: Parallel Documentation and add Unit Tests

[kshitij-maths]

Description

This PR resolves the issues found in the Parallel Reduced Order Model (ROM) module. It modernizes the ezyrb.parallel implementation to align with the current EZyRB (v1.3.3) API and introduces a dedicated unit test suite.

Fixes and Improvements

1. API Synchronization (#224)

• Attribute Update: Replaced deprecated .parameters and .snapshots calls with parameters_matrix and snapshots_matrix to match the core Database structure.
• Matrix Orientation: Implemented the "Transpose Sandwich" in the fit() and predict() methods. This ensures that data is passed to the POD and RBF engines in the correct $(n_features, n_samples)$ orientation.
• Signature Cleanup: Removed legacy arguments (scaler_red, database) from internal method calls that were causing TypeError during parallel task distribution.

2. Documentation & Quality (#225)

• Mathematical Docstrings: Added LaTeX-formatted documentation to the predict() method $(\mu \in \mathbb{R}^d \rightarrow u(\mu)).$
• Future-Proofing: Updated Delaunay imports from scipy.spatial.qhull to scipy.spatial to maintain compatibility with SciPy 2.0.
• New Test Suite: Created tests/test_parallel.py which covers:
  • test_initialization: Verifies proper object construction and attribute mapping.
  • test_fit: Confirms the parallel training workflow and generation of reduction modes.
  • test_predict_scalar: Tests the "Transpose Sandwich" logic for a single parametric input.
  • test_predict_db: Mathematically verifies accuracy by predicting a full database and comparing against original snapshots.
  • test_wrong_dimensions: Ensures the model correctly raises exceptions for mismatched input data.

Verification Results

pytest -v tests/test_parallel.py

• Result: 5/5 parallel tests passed.

Related Issues

This PR resolves issues #224 and #225