Speed up test_urlsplit_normalization (GH-26688)

diff --git a/Lib/test/test_urlparse.py b/Lib/test/test_urlparse.py
index dff9a8e..2f629c7 100644
--- a/Lib/test/test_urlparse.py
+++ b/Lib/test/test_urlparse.py
@@ -1079,7 +1079,8 @@
         hex_chars = {'{:04X}'.format(ord(c)) for c in illegal_chars}
         denorm_chars = [
             c for c in map(chr, range(128, sys.maxunicode))
-            if (hex_chars & set(unicodedata.decomposition(c).split()))
+            if unicodedata.decomposition(c)
+            and (hex_chars & set(unicodedata.decomposition(c).split()))
             and c not in illegal_chars
         ]
         # Sanity check that we found at least one such character