GIN fast insert

Hi there,

we present two variants of GIN fast insert patch, since we're not sure
what is a the best solution:

v0.28.1
- remove disable cost in gincostestimate
- per http://archives.postgresql.org/message-id/499466D2.4010808@sigaev.ru
   gingettuple could force cleanup of pending list if it got a lossy tidbitmap.
   If such cleanup occurs the gingettuple will rescan pending list again.

v0.28.2
- remove disable cost in gincostestimate
- per http://archives.postgresql.org/message-id/12795.1234379754@sss.pgh.pa.us
   AM can now have only one search method: amgettuple or amgetbitmap.
- GIN now has only amgetbitmap interface
--
Teodor Sigaev                                   E-mail: teodor@sigaev.ru
                                                    WWW: http://www.sigaev.ru/