Fix #2050: Symbols sections are appeared twice in the index page

Multibyte symbols are categorized to independent symbols section
different to single byte symbols.  This integrate them to a single
section.

CHANGES

@@ -65,6 +65,7 @@ Bugs fixed
 * C and C++, removed ``noindex`` directive option as it did
   nothing.
 * #7619: Duplicated node IDs are generated if node has multiple IDs
+* #2050: Symbols sections are appeared twice in the index page
 
 Testing
 --------

sphinx/environment/adapters/indexentries.py

@@ -98,9 +98,8 @@ class IndexEntries:
             for subentry in indexentry[1].values():
                 subentry[0].sort(key=keyfunc0)  # type: ignore
 
-        # sort the index entries; put all symbols at the front, even those
-        # following the letters in ASCII, this is where the chr(127) comes from
-        def keyfunc(entry: Tuple[str, List]) -> Tuple[str, str]:
+        # sort the index entries
+        def keyfunc(entry: Tuple[str, List]) -> Tuple[Tuple[int, str], str]:
             key, (void, void, category_key) = entry
             if category_key:
                 # using specified category key to sort
@@ -108,11 +107,16 @@ class IndexEntries:
             lckey = unicodedata.normalize('NFD', key.lower())
             if lckey.startswith('\N{RIGHT-TO-LEFT MARK}'):
                 lckey = lckey[1:]
+
             if lckey[0:1].isalpha() or lckey.startswith('_'):
-                lckey = chr(127) + lckey
+                # put non-symbol characters at the folloing group (1)
+                sortkey = (1, lckey)
+            else:
+                # put symbols at the front of the index (0)
+                sortkey = (0, lckey)
             # ensure a determinstic order *within* letters by also sorting on
             # the entry itself
-            return (lckey, entry[0])
+            return (sortkey, entry[0])
         newlist = sorted(new.items(), key=keyfunc)
 
         if group_entries:

tests/test_environment_indexentries.py

@@ -25,12 +25,14 @@ def test_create_single_index(app):
             ".. index:: ёлка\n"
             ".. index:: ‏תירבע‎\n"
             ".. index:: 9-symbol\n"
-            ".. index:: &-symbol\n")
+            ".. index:: &-symbol\n"
+            ".. index:: £100\n")
     restructuredtext.parse(app, text)
     index = IndexEntries(app.env).create_index(app.builder)
     assert len(index) == 6
     assert index[0] == ('Symbols', [('&-symbol', [[('', '#index-9')], [], None]),
-                                    ('9-symbol', [[('', '#index-8')], [], None])])
+                                    ('9-symbol', [[('', '#index-8')], [], None]),
+                                    ('£100', [[('', '#index-10')], [], None])])
     assert index[1] == ('D', [('docutils', [[('', '#index-0')], [], None])])
     assert index[2] == ('P', [('pip', [[], [('install', [('', '#index-2')]),
                                             ('upgrade', [('', '#index-3')])], None]),