refactor(generator): migrate univariate specs to pysatl-core distribtions

[Desiment]

Migrate pysatl_cpd/data/generator univariate sampling/specs from SciPy-style classes (NormalSpec, UniformSpec, etc.) to pysatl-core with UnivariateDistributionSpec, YAML support for optional parameterization, and UNURAN sampling; then update docs/notebooks and satisfy hooks/coverage.

Description of incoming changes:

• Replace only: remove old public univariate spec classes; no compatibility aliases.
• Use best-effort reproducibility; exact seeded reproducibility is not possible for core/UNURAN sampling now, so corresponding golden test marked as expecting to fail.
• Drop Student-t distribution entirely.
• Keep YAML support, including old kind: normal|uniform|exponential shapes translated into the new unified spec.
• Use pysatl-core family names/parameter names:
  • UnivariateDistributionSpec("Normal", "meanStd", mu=..., sigma=...)
  • UnivariateDistributionSpec("ContinuousUniform", "standard", lower_bound=..., upper_bound=...)
  • UnivariateDistributionSpec("Exponential", "scale", beta=...)

Detailied description

• Replaced univariate spec model in pysatl_cpd/data/generator/specs.py with UnivariateDistributionSpec.
• Updated YAML/config parsing in pysatl_cpd/data/generator/config.py:
  • old shapes kind: normal|uniform|exponential now map to UnivariateDistributionSpec
  • new generic kind: univariate with optional parametrization_name supported
  • student_t removed/rejected
• Switched univariate sampling in pysatl_cpd/data/generator/segments/sampling.py to pysatl-core + DefaultUnuranSamplingStrategy.
• Updated generator presets in pysatl_cpd/data/generator/presets.py to use unified specs.
• Updated exports in pysatl_cpd/data/generator/__init__.py to remove old univariate classes.
• Rewrote tests to use UnivariateDistributionSpec:
  • tests/unit/data/generator/test_specs.py
  • tests/unit/data/generator/test_config.py
  • tests/unit/data/generator/segments/test_sampling.py
  • tests/unit/data/generator/test_series.py
  • tests/unit/data/generator/test_dataset_generator.py
  • tests/unit/data/dataset/test_dataset_split.py
  • tests/regression/test_seeded_generators.py
• Marked seeded golden regression as xfail because pysatl-core UNURAN sampling lacks exposed seed control.
• Rewrote remaining Python doc examples from NormalSpec(...) style to UnivariateDistributionSpec(...) in multiple __init__.py files across the codebase.
• Updated notebooks:
  • notebooks/user_guide/02-generator-api.ipynb
  • notebooks/user_guide/04-core-api-visualization.ipynb
• Validation completed successfully:
  • poetry run pytest tests/unit/data/generator tests/unit/data/dataset/test_dataset_split.py tests/regression/test_seeded_generators.py → 97 passed, 1 xfailed
  • poetry run ruff check → passed
  • poetry run mypy → passed
  • poetry run pre-commit run --all-files → passed
  • Notebook JSON validation → passed

Key Decisions

• Use only UnivariateDistributionSpec as the public univariate spec API; remove NormalSpec, UniformSpec, ExponentialSpec, StudentTSpec.
• Keep multivariate normal sampling on NumPy; only univariate/independent-column univariate sampling moved to pysatl-core UNURAN.
• Dp not preserve old YAML univariate shapes by translating them to core-backed unified specs.
• Drop Student-t entirely per user request.
• Accept best-effort seeded behavior for core/UNURAN; exact seeded golden behavior is not preserved.