Issue #9425: fix setup_context() for non-ascii filenames

setup_context() replaces .pyc or .pyo filename suffix by .py, but it
didn't work if the filename contains a non-ascii character because the
function used the wrong unit for the length (number of characters
instead of the number of bytes).

With this patch, it uses unicode filenames instead of bytes filenames,
to fix the bug and to be fully unicode compliant.
diff --git a/Python/_warnings.c b/Python/_warnings.c
index dd7bb57..6067ce3 100644
--- a/Python/_warnings.c
+++ b/Python/_warnings.c
@@ -498,23 +498,21 @@
     *filename = PyDict_GetItemString(globals, "__file__");
     if (*filename != NULL) {
         Py_ssize_t len = PyUnicode_GetSize(*filename);
-        const char *file_str = _PyUnicode_AsString(*filename);
-            if (file_str == NULL || (len < 0 && PyErr_Occurred()))
-            goto handle_error;
+        Py_UNICODE *unicode = PyUnicode_AS_UNICODE(*filename);
 
         /* if filename.lower().endswith((".pyc", ".pyo")): */
         if (len >= 4 &&
-            file_str[len-4] == '.' &&
-            tolower(file_str[len-3]) == 'p' &&
-            tolower(file_str[len-2]) == 'y' &&
-            (tolower(file_str[len-1]) == 'c' ||
-                tolower(file_str[len-1]) == 'o'))
+            unicode[len-4] == '.' &&
+            Py_UNICODE_TOLOWER(unicode[len-3]) == 'p' &&
+            Py_UNICODE_TOLOWER(unicode[len-2]) == 'y' &&
+            (Py_UNICODE_TOLOWER(unicode[len-1]) == 'c' ||
+                Py_UNICODE_TOLOWER(unicode[len-1]) == 'o'))
         {
-            *filename = PyUnicode_FromStringAndSize(file_str, len-1);
-                if (*filename == NULL)
-                        goto handle_error;
-            }
-            else
+            *filename = PyUnicode_FromUnicode(unicode, len-1);
+            if (*filename == NULL)
+                goto handle_error;
+        }
+        else
             Py_INCREF(*filename);
     }
     else {