feat(search): lazily import dateparser

It can be expensive to import due to building thousands of regexps at
import time (see scrapinghub/dateparser#1181).

weblate/utils/search.py

@@ -12,7 +12,6 @@
 from operator import and_, or_
 from typing import Any, cast, overload
 
-from dateparser import parse as dateparser_parse
 from dateutil.parser import ParserError
 from dateutil.parser import parse as dateutil_parse
 from django.db import transaction
@@ -240,6 +239,21 @@ def convert_datetime(self, text, hour=5, minute=55, second=55, microsecond=0):
                 ),
             )
 
+        return self.human_date_parse(text, hour, minute, second, microsecond)
+
+    def human_date_parse(
+        self,
+        text: str,
+        hour: int = 5,
+        minute: int = 55,
+        second: int = 55,
+        microsecond: int = 0,
+    ) -> datetime | tuple[datetime, datetime]:
+        # Lazily import as this can be expensive
+        from dateparser import parse as dateparser_parse
+
+        tzinfo = timezone.get_current_timezone()
+
         # Attempts to parse the text using dateparser
         # If the text is unparsable it will return None
         result = dateparser_parse(text)