Add AL evaluation/benchmark papers (NeurIPS 2023, TMLR 2025, TMLR 2026)

[sten2lu]

Hi! Adding three papers on AL evaluation methodology:

1. [NeurIPS 2023] Navigating the Pitfalls of AL Evaluation — large-scale benchmark identifying 5 common evaluation flaws. Added to: NeuraIPS.md, Benchmarks section, deep_AL_criticism.md.

2. [TMLR 2025] nnActive — largest AL benchmark for 3D biomedical segmentation (~150k GPU hours, 8 methods × 4 datasets). Code: https://github.com/MIC-DKFZ/nnActive, Results: https://huggingface.co/nnActive. Added to: Benchmarks section, deep_AL_criticism.md.

3. [TMLR 2026] Finally Outshining the Random Baseline — simple effective AL solution consistently beating random in 3D medical imaging. Code: https://github.com/MIC-DKFZ/nnActive, Results: https://huggingface.co/nnActive. Added to: Benchmarks section, deep_AL_criticism.md.

Happy to adjust placement or format if needed.