move PREPARED_GEOM define in postgis_config.h

[strk]

The PREPARED_GEOM define is a configuration directive, so it'd be best placed in postgis_config.h. Currently it is in lwgeom_geos_prepared.h and failing to include it from an implementation file results in disabling PREPARED geometries w/out further build errors. See ticket #1165

[strk]

The only reason for PREPARED_GEOM to be undefined is for builds againts geos 3.1 Since we _require_ 3.2 now, there's no point in making it a configuration. I've moved the define in the implementation file, where it is used, with r7879.

@robe: worth checking it didn't slowdown again (you never know!)

[strk]

r7880 makes it even build (thanks Hudson!)

[robe]

Do you have debug code stuck somewhere? my regress is failing

Run tests: 74
Failed: 70

with stuff like this using r7880

--- loader/PointM-w.select.expected	Fri Sep  9 06:49:41 2011
+++ /tmp/pgis_reg_4740/test_2_out	Wed Sep 21 22:02:12 2011
@@ -1,3 +1,6 @@
+NOTICE:  [lwgeom_functions_basic.c:LWGEOM_asEWKT:2296] LWGEOM_asEWKT called.
+NOTICE:  [lwgeom_functions_basic.c:LWGEOM_asEWKT:2296] LWGEOM_asEWKT called.
+NOTICE:  [lwgeom_functions_basic.c:LWGEOM_asEWKT:2296] LWGEOM_asEWKT called.
 POINTM(0 1 3)
 POINTM(9 -1 -123)
 POINTM(9 -1 -3)

[robe]

disregard. For some reason I had enable-debug on my 9.1 configure. My 9.0 was fine. I suspect gremlins are changing my scripts since I don't recall changing my build script in past couple of days and certainly not adding in an enable-debug.

Though I thought we stripped notices from our regress tests results.

[strk]

It's most likely gremlins, see #1205