Change sys.maxunicode to 0xffff

[slozier]

Using 0xffff for sys.maxunicode is probably more accurate since our implementation of unicode is closer that of CPython 3.2.

Some Python code uses this value to check if they can use "wide" characters, for example, from docutils:

if sys.maxunicode >= 0x10FFFF: # "wide" build
    delimiters += '\U00010100\U00010101\U0001039f\U000103d0\U00010857'

Another example from pyyaml:

has_ucs4 = sys.maxunicode > 0xffff

In both these modules, we end up with errors about "invalid" character ranges in a regular expressions:

re.compile("[\U00010000-\U0010ffff]")

re.error: parsing "[-]" - [x-y] range in reverse order.