pgsql: Improve planner's estimation of the space needed for HashAgg

Log Message:
-----------
Improve planner's estimation of the space needed for HashAgg plans:
look at the actual aggregate transition datatypes and the actual overhead
needed by nodeAgg.c, instead of using pessimistic round numbers.
Per a discussion with Michael Tiemann.

Modified Files:
--------------
pgsql/src/backend/executor:
nodeAgg.c (r1.127 -> r1.128)
(http://developer.postgresql.org/cvsweb.cgi/pgsql/src/backend/executor/nodeAgg.c.diff?r1=1.127&r2=1.128)
pgsql/src/backend/optimizer/plan:
planner.c (r1.177 -> r1.178)
(http://developer.postgresql.org/cvsweb.cgi/pgsql/src/backend/optimizer/plan/planner.c.diff?r1=1.177&r2=1.178)
pgsql/src/backend/optimizer/util:
clauses.c (r1.186 -> r1.187)
(http://developer.postgresql.org/cvsweb.cgi/pgsql/src/backend/optimizer/util/clauses.c.diff?r1=1.186&r2=1.187)
pgsql/src/include/executor:
nodeAgg.h (r1.23 -> r1.24)
(http://developer.postgresql.org/cvsweb.cgi/pgsql/src/include/executor/nodeAgg.h.diff?r1=1.23&r2=1.24)
pgsql/src/include/optimizer:
clauses.h (r1.77 -> r1.78)
(http://developer.postgresql.org/cvsweb.cgi/pgsql/src/include/optimizer/clauses.h.diff?r1=1.77&r2=1.78)