Just build a custom alpha-numeric pattern for your purpose:
accents = [u'\u064b', u'\u064c', u'\u064d', u'\u064e', u'\u064f', u'\u0650', u'\u0651', u'\u0652']
alnum = r'([\w%s]+)' % re.escape(''.join(accents))
pattern = re.compile(alnum, re.UNICODE)
To find all tokens:
>>> test_str = "...foo" + ''.join(accents) + "...bar"
>>> test_str
'...fooًٌٍَُِّْ...bar'
>>> pattern.findall(test_str)
['fooًٌٍَُِّْ', 'bar']
>>> len(_)
2
Now to put a space between tokens and the rest:
>>> ' '.join(filter(None, pattern.split(test_str)))
'... fooًٌٍَُِّْ ... bar'
>>> len(_.split())
4