fix: ensure openai gpt-5.5 limits are exact

Author: rekram1-node

providers/openai/models/gpt-5.5.toml

@@ -22,8 +22,8 @@ cache_read = 1.00
 
 [limit]
 context = 1_050_000
-input = 920_000
-output = 130_000
+input = 922_000
+output = 128_000
 
 [modalities]
 input = ["text", "image", "pdf"]