Yield GIL in discovery ring computation to unblock uwsgi during reforge

The O(n²) greedy walk held the GIL for the entire computation, starving
uwsgi workers on the 2 vCPU box. time.sleep(0) in the inner loop
releases the GIL each iteration so request handling can interleave.

make_post_sell/models/shop.py

@@ -1,6 +1,7 @@
 import json
 import logging
 import threading
+import time
 import uuid
 
 from sqlalchemy import Column, BigInteger, Boolean, Unicode, UnicodeText, func
@@ -665,6 +666,8 @@ def compute_discovery_ring(shop):
                 score = len(accumulated_stems & p_stems) / len(union)
             else:
                 score = 0
+            # Yield GIL so uwsgi workers can serve requests
+            time.sleep(0)
 
             # Tiebreak: highest score, then newest, then ID
             if (score > best_score or