Major overhaul of large-object implementation, by Denis Perchine with

kibitzing from Tom Lane.  Large objects are now all stored in a single
system relation "pg_largeobject" --- no more xinv or xinx files, no more
relkind 'l'.  This should offer substantial performance improvement for
large numbers of LOs, since there won't be directory bloat anymore.
It'll also fix problems like running out of locktable space when you
access thousands of LOs in one transaction.
Also clean up cruft in read/write routines.  LOs with "holes" in them
(never-written byte ranges) now work just like Unix files with holes do:
a hole reads as zeroes but doesn't occupy storage space.
INITDB forced!

src/test/regress/expected/opr_sanity.out

@@ -482,8 +482,8 @@ WHERE p1.aggtransfn = p2.oid AND
           (p2.pronargs = 1 AND p1.aggbasetype = 0)));
   oid  | aggname | oid |   proname   
 -------+---------+-----+-------------
- 16984 | max     | 768 | int4larger
- 16998 | min     | 769 | int4smaller
+ 16996 | max     | 768 | int4larger
+ 17010 | min     | 769 | int4smaller
 (2 rows)
 
 -- Cross-check finalfn (if present) against its entry in pg_proc.

src/test/regress/expected/sanity_check.out

@@ -40,6 +40,7 @@ SELECT relname, relhasindex
  pg_index            | t
  pg_inherits         | t
  pg_language         | t
+ pg_largeobject      | t
  pg_listener         | t
  pg_opclass          | t
  pg_operator         | t
@@ -54,5 +55,5 @@ SELECT relname, relhasindex
  shighway            | t
  tenk1               | t
  tenk2               | t
-(44 rows)
+(45 rows)