Release v2.0.0: trial-based timing architecture

[ValentinGuigon]

Summary

Merges release/v2.0.0-prep into master for the neurodesign-plus 2.0.0 release: a source-breaking rework of the timing model around conceptual trials and modeled events, replacing the inherited event-only/ITI-based architecture.

Full details: CHANGES.md · MIGRATION_2.0.md

Highlights

• New trial-based timing model — trial_start_interval, post_event_interval, event_transition_interval, inter_trial_interval, rest_every_n_trials, rest_interval replace the old ambiguous ITI handling.
• Trial-template design modes — flat one-event shorthand, fixed complete-trial templates (trial_templates + trials), and probabilistic complete-template sampling (trial_templates + trial_template_probabilities + n_conceptual_trials), replacing event-level template truncation.
• Standardized workflow — Experiment.create_design() / create_manual_design() for construction, optimise() → selected_design(rank) for retrieval.
• RNG rewrite — NumPy SeedSequence/Generator throughout, removing global-state reseeding.
• Convergence reporting — patience-based early stopping (convergence=k), with generations_completed, stop_reason, optima exposed.
• Rewritten tutorials (tutorials/) covering the v2 workflow end-to-end; renamed/clarified validation scripts under validation/; improved public docstrings on Experiment and Optimisation.

Breaking changes

• Removed: stimuli_durations, conditional_ITI, order_keys/order_probabilities/order_length, all_stim_durations, t_pre/t_post, ITImodel/ITImin/ITImean/ITImax, direct .bestdesign reliance.
• Fixed-rule timing dicts now use "value" instead of "mean" for {"model": "fixed", ...} (uniform/exponential/gaussian rules are unaffected).
• See MIGRATION_2.0.md for the full v1 → v2 upgrade path.

Validation

• Full pytest suite passing (71 passed).
• Legacy-parity comparison against the original package retained under tests/test_legacy_parity.py.
• All maintained tutorial notebooks re-executed clean from fresh kernels.
• Version metadata synced to 2.0.0 across pyproject.toml and CITATION.cff.