"Can I Eat This?" — Benchmarking Whether AI Models Know Where Their Foraging Knowledge Ends