Guidelines for workflow definition

[frazane]

This PR introduces guidelines for extending the evalml workflow and enforces them. It's almost entirely renaming some rules and files so it should not break anything, but I would test anyway.

• Rename verif.smk → verification.smk and all verif_* rules to verification_* to match the {module}_ prefix convention
• Rename rules to follow {module}_{operation} pattern (e.g. prepare_checkpoint → inference_get_checkpoint, create_inference_venv → inference_create_venv)
• Rename scripts to mirror their calling rules (e.g. extract_baseline.py → data_extract_baseline.py, verif_single_init.py → verification_metrics.py)
• Remove vzarr_references.py (dead code)
• Add a "Workflow development guidelines" section to the README documenting rule/script naming, log paths, output layout, and wildcard conventions

[Louis-Frey]

Thanks Francesco, good Idea! I will test the configs and let you know.

[Louis-Frey]

So I tested all the configs, with the exception of forecasters-co1e.yaml (this was expected to fail), and before main was merged into this branch.

forecasters-ich1-oper-fixed.yaml produces some errors when initially run on the 2024 cases, but when I run it first on the 2025 case, and then on the 2024 cases again, it runs fine. I will take it low-prio on my list to look into that as well.

The other configs run without error. Good to go from my side.

[frazane]

forecasters-ich1-oper-fixed.yaml produces some errors when initially run on the 2024 cases, but when I run it first on the 2025 case, and then on the 2024 cases again, it runs fine. I will take it low-prio on my list to look into that as well.

Since this is the config that we use for the latest experiments, I would not merge this until this is fixed. What was the error?