Fix tokenizer: backslashes sometimes led to newline token generation

parso/python/tokenize.py

@@ -483,7 +483,7 @@ def tokenize_lines(lines, version_info, start_pos=(1, 0)):
                 break
             initial = token[0]
 
-            if new_line and initial not in '\r\n#':
+            if new_line and initial not in '\r\n\\#':
                 new_line = False
                 if paren_level == 0 and not fstring_stack:
                     i = 0