Change the apply trigger query cache to use a small footprint of the query
as the hash key. This key can be computed without doing any of the column
type lookups, making the query cache more efficient. If a prepared plan
is found, none of the catalog lookups are required at all.

Add a configuration option "apply_cache_size" to slon.conf.

Use custom wrappers around hash_any() and strcpm() to eliminate the use of md5.

Branch
------
master

Details
-------
http://git.postgresql.org/gitweb?p=slony1-engine.git;a=commitdiff;h=1ac9259fa2849de4d065c0d092945886f0a1aeb3

Modified Files
--------------
share/slon.conf-sample       |   21 +-
src/backend/slony1_funcs.c   |  838 +++++++++++++++++++++++++-----------------
src/backend/slony1_funcs.sql |   14 +-
src/slon/confoptions.c       |   12 +
src/slon/confoptions.h       |    2 +
src/slon/remote_worker.c     |   14 +-
src/slon/slon.c              |    4 +-
7 files changed, 541 insertions(+), 364 deletions(-)