Add trigger to accounting.cdr.start_time.

This statement is explicitely executed on both nodes separately and is
bypassing replication, to avoid having it replicated to the active node,
which will potentially lock the table for quite a while if it's big.

db_scripts/diff/13042_not_replicated.down

@@ -0,0 +1,2 @@
+SET sql_log_bin=0;
+ALTER TABLE accounting.cdr DROP KEY stime_idx;

db_scripts/diff/13042_not_replicated.up

@@ -0,0 +1,8 @@
+-- We explicitely don't replicate this statement in order
+-- to perform the query only on the corresponding standby
+-- node, and we make sure to execute it on both nodes.
+-- The reason is that this statement might take quite long
+-- and would lock the cdr table on the active node.
+
+SET sql_log_bin=0;
+ALTER TABLE accounting.cdr ADD KEY stime_idx(start_time);