Make pyrsistent._pmap doctests order-insensitive

[musicinmybrain]

This PR fixes #238 by removing all doctest assumptions about the particular hash values of str or bytes objects.

[musicinmybrain]

The doctests in pyrsistent._pset are still sensitive to the hash order of small integers. While this could change in theory, these have always hashed to themselves in CPython.

I would be happy to apply a similar adjustment there if desired, e.g.:

    >>> s = pset([1, 2, 3, 1])
    >>> s2 = s.add(4)
    >>> s3 = s2.remove(2)
    >>> s == {1, 2, 3}
    True

The doctests in pyrsistent._pbag are also sensitive to the hash order of small integers. These are a little more awkward to handle; I think something like this would be needed:

    >>> s = pbag([1, 2, 3, 1])
    >>> s2 = s.add(4)
    >>> s3 = s2.remove(1)
    >>> sorted(s)
    [1, 1, 2, 3]

Please let me know whether you would like me to make these changes or leave the doctests that depend only on small-integer hashing alone.

[tobgu]

Great, thanks!