Add ECGMultiLabelCardiologyTask

[jgong817]

Contributor: Jonathan Gong (jgong11@illinois.edu), Misael Lazaro (misaell2@illinois.edu), Sydney Robeson (sel9@illinois.edu)

Contribution Type: New Task

Description: ECGMultiLabelCardiologyTask is a standalone PyHealth task for multi-label ECG
classification. It is designed to operate on CardiologyDataset-style records
that reference paired PhysioNet-format waveform (.mat) and header (.hea) files. This task is based off of the work done by Nonaka and Seita (2021): https://proceedings.mlr.press/v149/nonaka21a/nonaka21a.pdf

For each visit record, the task:

1. Loads a 12-lead ECG signal from the waveform file.
2. Parses diagnosis codes and basic demographics (sex and age) from the
   corresponding header file.
3. Converts the configured diagnosis label set into a multi-hot target vector.
4. Segments the ECG into fixed-length sliding windows using the specified epoch
   length, shift, and sampling rate.
5. Produces model-ready samples containing the signal window, multilabel target,
   and associated visit/patient metadata.

This makes the task suitable for training and testing existing PyHealth models
on synthetic or dataset-backed ECG classification workflows.

Files to Review:

• pyhealth/tasks/ecg_classification.py - Main task implementation
• docs/api/tasks/pyhealth.tasks.ecg_classification.rst - Documentation file
• docs/api/tasks.rst - Updated task index
• tests/core/test_ecg_classification_task.py - Test cases with synthetic data
• pyhealth/examples/ecg_multilabel_mlp.py - Task ablation study