[utils] Fix sanitize_url regression with search queries

Commit 1baa0f5 introduced a regression which caused search queries such as
"ytsearch:search query" to be changed to "ytsearch:search%20query".
This arose when setting the default-search option to "ytsearch"
and when calling the YoutubeDL.extract_info method.

This occurred because sanitize_url, which doesn't always receive a URL,
performs a URL escape. Instead, this escape should be performed in
sanitized_Request, which always receives a URL.

test/test_utils.py

@@ -65,6 +65,7 @@ from youtube_dl.utils import (
     sanitize_filename,
     sanitize_path,
     sanitize_url,
+    sanitized_Request,
     expand_path,
     prepend_extension,
     replace_extension,
@@ -236,6 +237,10 @@ class TestUtil(unittest.TestCase):
         self.assertEqual(sanitize_url('httpss://foo.bar'), 'https://foo.bar')
         self.assertEqual(sanitize_url('rmtps://foo.bar'), 'rtmps://foo.bar')
         self.assertEqual(sanitize_url('https://foo.bar'), 'https://foo.bar')
+        self.assertEqual(sanitize_url('ytsearch:search query'), 'ytsearch:search query')
+
+    def test_sanitized_Request(self):
+        self.assertEqual(sanitized_Request('https://foo.bar/foo bar').get_full_url(), 'https://foo.bar/foo%20bar')
 
     def test_expand_path(self):
         def env(var):

youtube_dl/utils.py

@@ -2169,11 +2169,11 @@ def sanitize_url(url):
     for mistake, fixup in COMMON_TYPOS:
         if re.match(mistake, url):
             return re.sub(mistake, fixup, url)
-    return escape_url(url)
+    return url
 
 
 def sanitized_Request(url, *args, **kwargs):
-    return compat_urllib_request.Request(sanitize_url(url), *args, **kwargs)
+    return compat_urllib_request.Request(escape_url(sanitize_url(url)), *args, **kwargs)
 
 
 def expand_path(s):