[estebanzimanyi]

An aggregate whose PostgreSQL name a function also holds cannot take that
name in DuckDB, Spark or Flink, which hold one function per name: those
engines call tAnd the aggregate tAndAgg. MobilityDB states that name where
it states every other engine's name, in @altsqlfn, on the transition
wrapper of the aggregate beside its @sqlaggfn, a transition serving one
aggregate:

@sqlfn tAndTransition()
@sqlaggfn tAnd()
@altsqlfn tAndAgg()

There @altsqlfn names the aggregate, never the transition: wrapper_names
leaves out a wrapper stating @sqlaggfn, so no transition signature takes
an altSqlName, and attach_aggregate_alt_sql_names, run after
attach_aggregates, pairs the tag with the @sqlaggfn names by position and
sets altSqlName on every entry of the aggregates section of that
aggregate. A wrapper whose two lists differ in length, or an aggregate two
wrappers name differently, stops the catalog, as an unresolved function
alternative does.

Witness: over the MobilityDB branch sql/aggregates-keep-their-names, whose
transition wrappers state @altsqlfn (Tbool_tand_transfn: tAndAgg), master's
parser states an alternative name on 1018 function signatures against 864
over MobilityDB master: the 154 transition signatures take the aggregate's
name as their own, and no aggregate entry takes it.

Why: one tag and one field carry every name an engine takes in place of a
PostgreSQL one, so the bindings read altSqlName for an aggregate as they
do for a function, and need no rule of their own to append Agg.

Measured: over that branch the catalog states altSqlName on 154 aggregate
entries, the CREATE AGGREGATE count of the eleven aggregates whose name a
function also holds, and on 864 function signatures, the count master's
parser and this one both read over MobilityDB master 439dc9fbe3 (0
aggregate entries there, which carries no such tag); the test suite runs
558 tests, OK, the two new ones included.