[P8-01] Dogfood burnt.check() on 5+ real-world notebooks

[RedPandaMC]

Field	Value
ID	P8-01-dogfood
Phase	8 — Validation
Priority	high
Blocked by	—

Redesign Notes

Dogfooding now targets the new workflow:

1. burnt.start_session() → run Spark code → burnt.check() → review findings
2. Test across multiple environments: local pyspark, Databricks, EMR/Dataproc (if possible)
3. Focus on actionable advice quality, not dollar accuracy

Remove DLT-specific dogfooding from core — that belongs in burnt[databricks] testing.

Remaining Work

• Run burnt.check() on 5+ real-world notebooks
• Verify actionable advice is specific and correct
• Document compute time estimates vs actual Spark metrics