Speed-up character data loading

In Python 3, sys.maxunicode is 0x10FFFF, this caused slow loading, and
in fact it is not necessary in our code.

And clear some unused data to save memory use.

esprima/character.py

@@ -23,16 +23,14 @@
 
 from __future__ import absolute_import, unicode_literals
 
-import sys
-
 import unicodedata
 from collections import defaultdict
 
 from .compat import uchr, xrange
 
 # https://stackoverflow.com/questions/14245893/efficiently-list-all-characters-in-a-given-unicode-category
 U_CATEGORIES = defaultdict(list)
-for c in map(uchr, xrange(sys.maxunicode + 1)):
+for c in map(uchr, xrange(0x10000)):
     U_CATEGORIES[unicodedata.category(c)].append(c)
 UNICODE_LETTER = set(
     U_CATEGORIES['Lu'] + U_CATEGORIES['Ll'] +
@@ -82,6 +80,9 @@
 OCTAL_DIGIT = set(OCTAL_CONV.keys())
 HEX_DIGIT = set(HEX_CONV.keys())
 
+del U_CATEGORIES, UNICODE_LETTER, UNICODE_COMBINING_MARK
+del UNICODE_DIGIT, UNICODE_CONNECTOR_PUNCTUATION
+del DECIMAL_CONV, OCTAL_CONV, HEX_CONV
 
 class Character:
     @staticmethod