fix(test): correct prompt_tokens in test_string_cost_values to match token details

The test was failing because prompt_tokens (1000) didn't match the sum of
detail tokens (700 + 100 + 200 + 150 = 1150). This triggered the double-counting
detection in generic_cost_per_token, which recalculated text_tokens as 550
instead of 700, causing the cost assertion to fail.

Fix:
- prompt_tokens: 1000 -> 1150
- total_tokens: 1500 -> 1650
- Added comment explaining the token sum requirement

Fixes: litellm_mapped_tests_litellm_core_utils CI failure

tests/test_litellm/litellm_core_utils/llm_cost_calc/test_llm_cost_calc_utils.py

@@ -313,10 +313,12 @@ def test_string_cost_values():
     }
 
     # Test usage with various token types
+    # Note: prompt_tokens must equal the sum of all detail tokens to avoid double-counting adjustment
+    # text_tokens(700) + audio_tokens(100) + cached_tokens(200) + cache_creation_tokens(150) = 1150
     usage = Usage(
-        prompt_tokens=1000,
+        prompt_tokens=1150,
         completion_tokens=500,
-        total_tokens=1500,
+        total_tokens=1650,
         prompt_tokens_details=PromptTokensDetailsWrapper(
             audio_tokens=100, cached_tokens=200, text_tokens=700, image_tokens=None, cache_creation_tokens=150
         ),