forge-stream sync_elapsed_streams() is still O(n) even after migration to persistent storage — now reads persistent storage in a loop

[Austinaminu2]

Summary

sync_elapsed_streams() has been updated to use persistent() storage for stream reads, but is still O(n):

while stream_id < next_id {
    let maybe_stream: Option<Stream> = env.storage().persistent().get(&DataKey::Stream(stream_id));
    // ...
    stream_id += 1;
}

Persistent storage reads are metered the same as instance storage reads. With 1000 streams this still reads 1000 persistent storage entries per get_active_streams_count() call. The migration to persistent storage solved the size limit issue but not the CPU cost issue.

Tasks

• Remove the O(n) scan entirely from sync_elapsed_streams()
• Maintain ActiveStreamsCount purely via increment on create_stream() and decrement on cancel_stream() and stream expiry during withdraw()
• Add expiry detection in withdraw(): if now >= stream.end_time && stream.counted_active, decrement and set counted_active = false
• Add a test with 20 streams where half have elapsed, verifying get_active_streams_count() is correct without a full scan

Labels: bug, forge-stream, performance, critical