[StevenSong]

Alot of things drifted between initial experimentation, initial paper submission, rebuttal experiments, and paper resubmission. This PR aims to clean all that up and also adds in some extended experiments we're conduction. A high level list of changes is below:

• add protossl/models/_protoecgnet.py and protossl/protoecgnet_trainer.py which enables multiarm ProtoECGNet (1D-global, 2D-global, 2D-partial arms), a fusion classifier, a fusion-embedding stage, tuning, and bootstrapped evaluation, all trained over HEEDB. Configs, scripts, and notebooks are under large-user-study/.
• adds heedb_split_type to split HEEDB other than by year, and drop_blank_overread to drop ECGs with no physician overread.
• add USE_HEEDB_150 env var toggle to switch to the 150-label HEEDB target set. Adds pretraining configs and scripts for ProtoSSL and supervised prototypes on this set
• add BlackboxContraster model and train-contraster stage for baseline PCLR pretraining over HEEDB. Adds PCLR embedding extraction and a logistic-regression probe on those embeddings (5-3-run-pclr-logreg.sh, queue-pclr-logreg.sh).
• add data_wrapper_kwargs to LitData in trainer for the contrastive wrapper datasets, and data_kwargs now goes to the base dataset
• rename experiments to match the paper: labsup-proto/prosup → supproto and pass → protossl across scripts and run directories
• better modularization for ECG waveform plotting (full and segment views, baseline-wander removal, axis layout).
• cleanup of README (new paper title, bibtex citation, repo layout table, data sources, steps to reproduce, experiment naming scheme, environment variables, license)