fix: support non-Latin text in InMemoryMemoryService search

Fixes #5501. `_extract_words_lower` used `[A-Za-z]+` regex which
only matched ASCII letters, silently discarding Japanese, Chinese,
Korean, Cyrillic and other non-Latin characters.

Change to `\w+` with `re.UNICODE` to match all Unicode word characters.

Author: weiguangli-io

src/google/adk/memory/in_memory_memory_service.py

@@ -39,7 +39,7 @@ def _user_key(app_name: str, user_id: str) -> str:
 
 def _extract_words_lower(text: str) -> set[str]:
   """Extracts words from a string and converts them to lowercase."""
-  return set([word.lower() for word in re.findall(r'[A-Za-z]+', text)])
+  return set([word.lower() for word in re.findall(r'\w+', text, re.UNICODE)])
 
 
 class InMemoryMemoryService(BaseMemoryService):