Describe the feature

The op-proofs ExEx runs an automated prune_task to enforce the configured proofs_history_window (e.g., 7 days). Currently, this task starts immediately when the node launches.

If a node operator drastically reduces the retention window in the configuration (e.g., changing from 30 days to 7 days), the pruner will attempt to delete millions of blocks of historical data in the first run.

This causes:

1. Write Locks: The massive delete transaction locks the MDBX database tables for an extended period.
2. Performance Degradation: The I/O storm from deleting millions of keys stalls the main execution pipeline, potentially causing the node to fall out of sync.

Solution
We should implement a "Safety Threshold" check during the ExEx initialization phase, before the background pruner is spawned.

Additional context

No response