Add BulkRNABert model for cancer prognosis

[gwicho38]

Summary

This PR adds the BulkRNABert model for cancer prognosis tasks based on the paper by Gélard et al. (2025).

Key additions:

• BulkRNABertLayer: Transformer encoder layer for gene expression data
• BulkRNABert: Main model for cancer type classification (33 TCGA cancer types)
• BulkRNABertForSurvival: Model for survival prediction using Cox proportional hazards loss
• compute_c_index: Utility function for computing concordance index
• Comprehensive unit tests

Paper reference: Gélard et al. (2025) "BulkRNABert: Cancer prognosis from bulk RNA-seq based language models"

• Paper: https://www.biorxiv.org/content/10.1101/2024.06.13.598798
• Model: https://huggingface.co/InstaDeepAI/BulkRNABert

Implementation Details

The implementation follows PyHealth's BaseModel pattern and provides:

• Multiclass classification for 33 TCGA cancer types
• Binary classification support
• Survival prediction with Cox partial likelihood loss
• C-index metric computation for survival evaluation

Test Plan

• [x] Unit tests for BulkRNABertLayer encoder
• [x] Unit tests for BulkRNABert classification model
• [x] Unit tests for BulkRNABertForSurvival model
• [x] Unit tests for compute_c_index function
• [x] All tests pass locally

Context

This contribution was developed as part of a CS 598 DLH (Deep Learning for Healthcare) course project reproducing the BulkRNABert paper results.