input.Makefile: Use new sort_filenames to putmultiple numbered sources in the correct order, dealing correctly with embedded numbers that aren't padded with leading zeros

Added sort_filenames to sort a list of filenames, comparing embedded numbers numerically instead of lexicographically

schemas/postgresql.conf: Decreased shared_buffers again because 4000MB wasn't enough less than 4GB SHMMAX

schemas/postgresql.conf: Expressed shared_buffers in MB, since decimal GB doesn't seem to work anymore on 9.1

schemas/postgresql.conf: Decreased shared_buffers to 3.9GB, slightly less than SHMMAX

schemas/postgresql.conf: Optimized again using same changes as were applied to 8.4 version

schemas/postgresql.conf: Replaced with original 9.1 version

schemas/postgresql.conf: Optimized using analogous settings as postgresql.nimoy.conf

inputs/REMIB/src/nodes.make: Don't abort entire import on empty response, because an empty response is also returned for nodes that are temporarily down, not just nodes that don't exist (assumed to be after the highest numbered node). Instead, stop import after 150 nodes if user did not specify an explicit # nodes.

inputs/REMIB/src/nodes.make: Abort prefix on empty response using break, rather than just done = True, to avoid running any more code except the finally block. Moved metadata row validation outside metadata row retrieval try-except block.