basically by caching them.
- improve LRUCache so that size management only proceeds
in one thread, add a real get() method

logic to some degree in DefaultExecutionContext.  In particular
we are removing post_insert() which doesn't appear to be used
based on a survey of prominent third party dialects.  Callcounts
aren't added to existing execute profiling tests and inserts might be
a little better.
- simplify the execution_options join in DEC.  Callcounts don't
appear affected.

for memoization on a class that uses slots.
- apply many more __slots__.  mem use for nova now at 46% savings

improved in conjunction with loader directives such as
:func:`.joinedload` and :func:`.contains_eager` such that if
two :meth:`.PropComparator.of_type` modifiers of the same
base type/path are encountered, they will be joined together
into a single "polymorphic" entity, rather than replacing
the entity of type A with the one of type B.  E.g.
a joinedload of ``A.b.of_type(BSub1)->BSub1.c`` combined with
joinedload of ``A.b.of_type(BSub2)->BSub2.c`` will create a
single joinedload of ``A.b.of_type((BSub1, BSub2)) -> BSub1.c, BSub2.c``,
without the need for the ``with_polymorphic`` to be explicit
in the query.
fixes #3256

:class:`.Query` object offers dramatic speed improvements when
fetching large numbers of column-oriented rows.
fixes #3176

view.  So copy collections.OrderedDict and use MutableMapping to set up
keys, items, values on our own OrderedDict.

Conflicts:
	lib/sqlalchemy/engine/base.py