DVC Pipelines: Reproducible ML Workflows

Day 24 of 100 Days of MLOps. Running make_dataset.py then clean.py then train.py by hand is fragile and forgetful. A DVC pipeline declares your stages, their inputs and outputs in dvc.yaml — so DVC can rebuild the whole chain, in order, and skip any stage whose inputs haven't changed. It's make (Day 9) supercharged with data-awareness. Runs 100% locally on any OS.