init_sequence spill to hash table

Here http://www.postgresql.org/message-id/24278.1352922571@sss.pgh.pa.us there
was some talk about init_sequence being a bottleneck when many sequences
are used in a single backend.

The attached I think implements what was talked about in the above link
which for me seems to double the speed of a currval() loop over 30000
sequences. It goes from about 7 seconds to 3.5 on my laptop.

I thought I would post the patch early to see if this is actually wanted
before I do too much more work on it.

My implementation maintains using the linear list for sequences up to a
defined threshold (currently 32) then it moves everything over to a
hashtable and free's off the list.

A more complete solution would contain regression tests to exercise the
hash table code.

I know there is a desire to move sequences over to a single table still,
but I see this as a separate patch and storing current values in a hash
table for each backend should still be a win even if/when the single table
stuff gets implemented.

Regards

David Rowley