Handle benchmark errors gracefully instead of panicking

[qdrant-cloud-bot]

Summary

• Replace .unwrap() on run_benchmark() result in main() with proper error handling that prints the error and exits with code 1
• Previously, when all upsert retries were exhausted (e.g. during chaos testing with frequent node kills + resharding), the error propagated to main() and hit .unwrap(), producing an ugly panic with stack trace instead of a clean exit

Context

Observed on qdrant-chaos-testing-three (2026-03-25 17:00–19:00 UTC): the combination of kill-random-node cron (~every 10 min), resharding, and rolling updates caused upsert timeouts. After exhausting all 5 retries, bfb panicked with:

called `Result::unwrap()` on an `Err` value: Error in the response: The operation was cancelled Timeout expired

Test plan

• cargo check passes
• Run bfb against a stopped/unreachable cluster and verify it prints a clean error message and exits with code 1 instead of panicking

Made with Cursor