Remote test artifacts accumulate without cleanup

[sdebruyn]

Problem

When using --remote for Spark test execution, each run creates a directory under dbt-remote-runs/artifacts/{run_id}/ on the lakehouse. These are never cleaned up automatically, so storage usage grows unbounded over time.

Proposed solution

Add a cleanup mechanism — either:

1. A --remote-cleanup flag that prunes artifacts older than N days
2. Automatic cleanup at the start of each run (delete artifacts older than 7 days)
3. A separate maintenance script

Context

This was identified during code review of PR #208. The current design correctly isolates artifacts per run for concurrency safety, but has no garbage collection.