Add minibenchmark notebook

[Comp-era]

Summary

Adds an interactive Jupyter notebook that walks through the protein–peptide benchmarking workflow end to end

Changes

• notebook/HADDOCK3_protein_peptide_small_benchmark.ipynb — new notebook
• README.md — new "Interactive notebook" section under Quick Start
• versions.env — bump haddock3 to 2026.8.0, haddock-runner to v4.1.0

[amjjbonvin]

A few points on the notebook:

1. Would it work running it directly on a Mac?(I see you are pulling a x86 linux version of the runner)

2. When referring/describing files (like the config file for the test benchmark), do provide their full path so that users can find them

3. For rigidbody, the sampling should be ideally a multiple of the number of core for efficiency.

I did run the notebook on colab, the benchmark completed within 5 minutes. Nice! Could not test the EGI notebooks environment because of a current EGI SSO issue.

[Comp-era]

Thanks for the review — all three addressed:

Mac support (76c3575): the install cell now detects the platform (platform.system()/platform.machine()) and picks the matching haddock-runner release — universal-apple-darwin on macOS (covers Intel and Apple Silicon in one binary), x86_64/aarch64-unknown-linux-musl on Linux. No longer hardcoded to x86_64 Linux.

Full file paths (d548183): every file/config reference in the notebook prose now uses the full repo-relative path (e.g. docking_benchmarks/protein_peptide/unbound_true_interface_act-pass-test.yaml, docking_benchmarks/protein_peptide/stage_labels.json) instead of bare filenames.

Sampling as a multiple of ncores (bf0a917): rigidbody.sampling in the test scenario is now 40 (was 30), a clean multiple of ncores: 4. Also added a note in the notebook to check available cores and size ncores/sampling accordingly.

Also ran it on EGI notebooks — completed fine.