Keep demo sample data out of the volume students upload to

When a Databricks demo notebook auto-generates sample CSVs into a UC volume, give it its own volume (e.g. demo_sample) rather than the one students will upload real competition files into. Shared file names (news.csv here) silently collide: the generator's 'skip if exists' guard then makes the demo read the real file and break on its columns, or the student's upload overwrites the sample. Parameterize catalog/schema/volume as dbutils widgets and wire them to bundle variables so jobs and CD can point elsewhere. Also: ai_classify happily labels rumor/decoy headlines as real disruptions, which is a useful teaching point rather than a bug.