engine: room for 128 candidates

32 hid the tail of common readings: きょう has 41 kanji, 구 has 352
hanja. 128 covers every kanji entry and the hanja dictionary now keeps
that many per reading.

tests/mkhanja_test.py

@@ -128,12 +128,12 @@ class MkhanjaTest(unittest.TestCase):
         self.assertIn("한\t", result.stdout)
 
     def test_generator_keeps_runtime_candidate_limit(self):
-        rows = "".join(f"{chr(0x4E00 + n)}\t한\n" for n in range(33))
+        rows = "".join(f"{chr(0x4E00 + n)}\t한\n" for n in range(129))
         result = run(GENERATE, self.source(rows))
         self.assertEqual(result.returncode, 0, result.stderr)
         candidates = result.stdout.rstrip().split("\t", 1)[1].split()
-        self.assertEqual(len(candidates), 32)
-        self.assertEqual(candidates[-1], chr(0x4E00 + 31))
+        self.assertEqual(len(candidates), 128)
+        self.assertEqual(candidates[-1], chr(0x4E00 + 127))
 
 
 if __name__ == "__main__":