port the tests to unittest and add a few new tests.

Makes test_xdrlib pass.

Also add xdrlib._test into the test suite.

From SF patch #852334.

giving it a slight facelift

Use != instead of <> since <> is documented as "obsolescent".
Use "is" and "is not" when comparing with None or type objects.

longs where the top bit is set.  First, change the masks so that they
are `L' longs, otherwise the sign bits will get propagated to the
result.  Next, do not coerce to int before sending to pack_uint()
otherwise Python will generate an OverflowError.  Here is a test
program that fails without the patch, but now succeeds:

import xdrlib

addr = (132, 151, 1, 71)
uint = 0L
for a in addr:
    uint = (uint << 8) | a

ulong64 = uint << 32

p = xdrlib.Packer()
p.pack_uhyper(ulong64)
buf = p.get_buffer()
u = xdrlib.Unpacker(buf)
ulong64prime = u.unpack_uhyper()

if ulong64 == ulong64prime:
    print 'okay'
else:
    print 'bogus'

print ulong64, ulong64prime

    raise ConversionError, msg

where `msg' is passed as the argument to the constructor.

data formats.  The _xdr module is no longer used, since struct
supports the required IEEE floats and doubles.

(I have one doubt about not using _xdr.  The struct module doesn't
handle Inf, NaN and gradual underflow correctly.  If the _xdr module
does these things better, it may still have a (small) competitive
advantage.  On the other hand, since not all platforms support IEEE
floating point, it's not clear that it would be a good idea to ever
transfer Inf or NaNs.  Gradual underflow can be fixed in the struct
module.