Fix urlparse schemaless-behaviour on Python 3.9+ (#33289)

The urlparse stdlib call behaves differently on Python 3.9+ for
schemaless-URLs. It parses string hostname into a schema instead
of netloc. The issue is still open and discussed

* psf/requests#6455

and apparently it will not be solved, so we need to workaround it
if we still want to support legacy, schemaless URLs to be accepted
by Elasticsearch handler.

GitOrigin-RevId: dd73a0bffa6c4de93a2dd8dc4460b64aedc51255

airflow/providers/elasticsearch/log/es_task_handler.py

@@ -143,7 +143,9 @@ def format_url(host: str) -> str:
         parsed_url = urlparse(host)
 
         # Check if the scheme is either http or https
-        if not parsed_url.scheme:
+        # Handles also the Python 3.9+ case where urlparse understands "localhost:9200"
+        # differently than urlparse in Python 3.8 and below (https://github.com/psf/requests/issues/6455)
+        if parsed_url.scheme not in ("http", "https"):
             host = "http://" + host
             parsed_url = urlparse(host)