Update for brain decoding demo

[zaRizk7]

• Adds load_data function and data manifests to load pre-computed functional connectivity directly from Google Drive using gdown.
• Adds workaround for pandas import error due to numpy 2.0 by using --user flag during pip install and prepends site-packages when a Colab runtime is detected.
• Default config in config.py now aims to reproduce intra-site evaluation in (Kunda et al., TMI 2022) with logistic regression and base.yml is used to optimize runtime in Colab.
• Reduce NUM_SEARCH_ITER and NUM_SOLVER_ITER to allow quicker runtime in Colab for base.yml. Latest estimated runtime: 11 minutes 30 seconds.
• Uses HCP-ICA atlas with 32 ROIs as default for base.yml, ensuring quicker runtime.
• More detailed explanation regarding experiment settings in section markdowns.

Note: Once the PR is approved, the Colab handle needs to be updated to clone the main branch instead of brain-decoding branch.

[Copilot]

Pull Request Overview

This PR enhances the brain decoding tutorial by adding a generic data-loading utility, updating configuration defaults for faster Colab runs, and including manifests for downloading precomputed data from Google Drive.

• Introduces load_data function and JSON manifests for automated data fetch
• Updates default experiment settings (atlas, search/solver iterations, random state)
• Refines preprocessing and parsing modules exports

Reviewed Changes

Copilot reviewed 9 out of 9 changed files in this pull request and generated 1 comment.

Show a summary per file

File	Description
tutorials/brain-disorder-diagnosis/preprocess.py	Renamed and exported preprocessing functions via __all__
tutorials/brain-disorder-diagnosis/parsing.py	Exported compile_results and adjusted module exports
tutorials/brain-disorder-diagnosis/manifests/atlas.json	Added manifest entries for atlas directories
tutorials/brain-disorder-diagnosis/manifests/abide.json	Added manifest entries for ABIDE connectivity and phenotypes files
tutorials/brain-disorder-diagnosis/experiments/base.yml	Switched default atlas to hcp-ica, reduced search iterations, set random state
tutorials/brain-disorder-diagnosis/data.py	New load_data implementation with download helpers for manifests
tutorials/brain-disorder-diagnosis/config.py	Changed default data path, expanded and restructured config options

Comments suppressed due to low confidence (2)

tutorials/brain-disorder-diagnosis/parsing.py:9

• [nitpick] This reassigns MODEL and shadows the original list. Consider using a distinct name (e.g., MODEL_DISPLAY_NAMES) for the mapping to avoid confusion.

MODEL = {model: " ".join(model.split("_")).title() for model in MODEL}

tutorials/brain-disorder-diagnosis/config.py:11

• Several previously supported dataset options (e.g., BANDPASS, GLOBAL_SIGNAL_REGRESSION) were removed in this PR. This is a breaking change—consider deprecating them or updating downstream code accordingly.

_C.DATASET.PATH = DEFAULT_DIR