Quartz `CTFont` doesn't handle Python 3 bytes to str properly

[corranwebster]

See the following, which is the result of calling str on a bytes (Python 2 idiom) instead of explicitly using a codec.

In [2]: from kiva.quartz import CTFont                                          

In [3]: CTFont.default_font_info.names()                                        
Out[3]: 
["b'Kannada MN'",
 "b'Malayalam Sangam MN'",
 "b'Noto Sans Kannada'",
 "b'Beirut'",
 "b'DIN Alternate'",
 "b'PT Sans Caption'",
 "b'Devanagari MT'",
 "b'PingFang TC'",
 "b'Courier'",
 ... # etc
 "b'Gill Sans'"]

Culprit is this method:

enable/kiva/quartz/CTFont.pyx

Lines 17 to 27 in bb2c358

	cdef object _cf_string_to_pystring(CFStringRef cf_string):
	cdef char* c_string
	cdef char buf[256]
	c_string = CFStringGetCStringPtr(cf_string, kCFStringEncodingMacRoman)
	if c_string == NULL:
	success = CFStringGetCString(cf_string, buf, 256,
	kCFStringEncodingMacRoman)
	retval = str(buf)
	else:
	retval = str(c_string)
	return retval