Unlike Scheme where exactness is implemented as taints, the Python
implementation associated exactness with data types.  This created
inheritance issues (making an exact subclass of floats would result
in the subclass having both an explicit Exact registration and an
inherited Inexact registration).  This was a problem for the
decimal module which was designed to span both exact and inexact
arithmetic.  There was also a question of use cases and no examples
were found where ABCs for exactness could be used to improve code.
One other issue was having separate tags for both the affirmative
and negative cases.  This is at odds with the approach taken
elsewhere in the Python (i.e. we don't have an ABC both Hashable
and Unhashable).

across platforms:  it should now raise OverflowError on all
platforms.  (Previously it raised OverflowError only on
non IEEE 754 platforms.)

Also fix the (already existing) test for this behaviour
so that it actually raises TestFailed instead of just
referencing it.

Closes issue #2055. Thanks Giampaolo Rodola.

Closes issue #1960. Thanks Giampaolo Rodola.

Patch from GHOP 294 by David Marek.

Done as GHOP 238 by Josip Dzolonga.

Thanks Benjamin Peterson for the patch.

(Exploit at request.)

Some buildbots were consistently failing (e.g., amd64).
Also remove a couple of semi-colons.

string.

optional: it defaults to the traceback of the exception that is currently
being handled.