feat(datasets): pooch-based fetcher + migrate sample data to HypypData (#260)

* feat(datasets): add pooch-based fetcher and migrate sample data to HypypData

Removes all sample data from the HyPyP repo and replaces hardcoded paths
with a clean `hypyp.datasets` API backed by pooch, mirroring MNE's
dataset convention (~/mne_data/HypypData/).

- Add `hypyp/datasets.py`: registry + fetch helpers for EEG (.fif),
  EEGLAB (.set/.fdt), fNIRS (.snirf, Homer2/NIRX), XDF, and DCARE/FCS01
- Add `pooch>=1.8` dependency to pyproject.toml
- Remove all files from `data/` (now served from ppsp-team/HypypData)
- Update .gitignore to ignore data/, tutorial/results/
- Refactor `fnirs.DataBrowser` to use MNE-style cache instead of
  __file__-based root path; pass path= to pooch.retrieve()
- Fix `plots.py`: guard y_cuts against None in plot_wavelet_transform_weights
- Update all 8 tutorial notebooks to use hypyp.datasets
- Update tests: conftest.py, test_fnirs.py, test_xdf.py to use datasets
- Add test_datasets.py (20 tests, cold-cache verified)
- Add scripts/run_notebooks.sh for sequential notebook execution

Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>

* chore(tutorial): remove wavelet_wtc_in_depth notebook

Contains hardcoded paths to private IMAGINE lab data (/home/pfortin/)
that cannot be replaced with public HypypData samples. Not suitable
for a public tutorial.

Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>

---------

Co-authored-by: Claude Sonnet 4.6 <noreply@anthropic.com>

Author: Ramdam17

.gitignore

@@ -141,3 +141,13 @@ _build/
 # Preprocessed and results data
 data/results/
 private/
+
+# Tutorial outputs — generated locally, not committed
+tutorial/results/
+
+# Sample data — moved to HypypData repo (https://github.com/ppsp-team/HypypData)
+# Only data/Basehead.obj is kept (listed as package_data in pyproject.toml)
+data/*.fif
+data/EEGLAB/
+data/NIRS/
+data/XDF/