object bypass bit sanity for container contents

Noticed when adding a 'tip container' choice to item-actions for
context sensitive inventory (update pending).  Putting items into a
container with menustyle traditional and then takiing them out with
the #tip command while 'sanity_check' is On would produce warnings
once they were on the floor.

askchain() uses object bypassing to be able to cope with multi-drop
potentially changing invent, and it tried to reset that when done.
But it did so with the original object list (invent in this case)
and that doesn't reset individual objects that have been moved to
any other list.  The between-turn resetting of bypass bits wasn't
doing so for container contents.  The sanity check wasn't--still
isn't--checking those either, so it wasn't noticeable while items
were still inside the container.  But taking them out with #tip
doesn't touch any bypass bits, so between-turn reset isn't triggered
and the items that came out of the container with bypass set
continued to have it set while on floor.  sanity_check complained.

Change clear_bypasses() to handle container contents, and change
askchain() to call it instead of just clearing bypasses for whatever
is left of its input chain.  (The latter probably isn't necessary
now that the between-turn cleanup deals with contents.)

doc/fixes3-7-0.txt

@@ -877,6 +877,9 @@ restoring while attached ball or chain is on floor in a breach of a shop wall
 	might trigger an impossible about being positioned too far from hero
 don't try to catch up for lost time for shop damage repair in restdamage()
 	called from getlev(); let normal shopkeeper movement take care of it
+putting objects into a container with menustyle=traditional and then taking
+	them back out with #tip would result in complaints about obj bypass
+	bit being set if sanity_check was On
 
 
 Fixes to 3.7.0-x Problems that Were Exposed Via git Repository