Rationalize support for EMPTY

[pramsey]

Apply the rules for EMPTY geometries as laid out in DevWikiEmptyGeometry to all geometry functions.

[robe]

Just itemizing some of these cases as I come across them in tests.

SELECT ST_Distance_Sphere(ST_GeomFromText('POLYGON EMPTY',4326),ST_GeomFromText('POINT(1 3)',4326))

Gives error: geometry_distance_spheroid: Only point/line/polygon supported.

Though curiously (which I thought they shared some code)

SELECT ST_Distance(geography(ST_GeomFromText('POLYGON EMPTY',4326)),geography(ST_GeomFromText('POINT(1 3)',4326)));

Returns null as expected

[strk]

With r8382 I fixed handling of EMPTY by ST_Buffer and put the regression test in tickets.sql, but I guess it'd make more sense to have an empty.sql testcase testing each and every line from the wiki page.

Does anyone feel like stubbing it ? Or I'll do it.

[strk]

r8408 inaugurates a regression test for "empty"

[strk]

#692 is about the spatial predicates portion of this

[strk]

This is completed as of r8415