to transient objects with attributes unset would leak NEVER_SET,
and negated_contains_or_equals would do so for any transient
object as the comparison used only the committed value.
Repaired the NEVER_SET cases, fixes #3371, and also made
negated_contains_or_equals() use state_attr_by_column() just
like a non-negated comparison, fixes #3374

NEVER_SET symbol could easily leak into a lazyload query, subsequent
to the flush of a pending object.  This would occur typically
for a many-to-one relationship that does not use a simple
"get" strategy.   The good news is that the fix improves efficiency
vs. 0.9, because we can now skip the SELECT statement entirely
when we detect NEVER_SET symbols present in the parameters; prior to
:ticket:`3061`, we couldn't discern if the None here were set or not.
fixes #3368

same target more than once; it implicitly dedupes only in the case of
a relationship join, and due to :ticket:`3233`, in 1.0 a join
to the same table twice behaves differently than 0.9 in that it no
longer erroneously aliases.   To help document this change,
the verbiage regarding :ticket:`3233` in the migration notes has
been generalized, and a warning has been added when :meth:`.Query.join`
is called against the same target relationship more than once.
fixes #3367

determining remote side with semi-self-referential (e.g. two joined
inh subclasses referring to each other), non-simple join conditions
such that the parententity is taken into account and can reduce the
need for using the ``remote()`` annotation; this can restore some
cases that might have worked without the annotation prior to 0.9.4
via :ticket:`2948`. fixes #3364

:meth:`.Session.begin_nested` operations would fail to propagate
the "dirty" flag for an object that had been updated within
the inner savepoint, such that if the enclosing savepoint were
rolled back, the object would not be part of the state that was
expired and therefore reverted to its database state.
fixes #3352

FROM clauses when using the :meth:`.Query.update` or
:meth:`.Query.delete` methods; instead of silently ignoring these
fields if methods like :meth:`.Query.join` or
:meth:`.Query.select_from` has been called, an error is raised.
In 0.9.10 this only emits a warning.
fixes #3349
- don't needlessly call _compile_context() and build up a
whole statement that we never need.  Construct QueryContext
as it's part of the event contract, but don't actually call upon
mapper attributes; use more direct systems of determining the
update or delete table.
- don't realy need _no_select_modifiers anymore

commit phase of the session, which without it could cause
a "dictionary changed size during iter" error if garbage collection
interacted within the process.   Change was introduced by

between fully and match on those, rather than trying to
compare selectables; fixes #3347