Implement --slave-abort-blocking-timeout

Request from Monty, from a user:

Implement an option --slave-abort-blocking-timeout. When this is set lower than --lock-wait-timeout and a slave replicating query (like an ALTER TABLE) is waiting for more that @@slave_abort_blocking_timeout blocked on a user query, then that query is killed to allow the slave to continue.

This is intended to solve a problem, where there is a long-running SELECT for example running on a slave, and a DDL being replicated needs to wait for that SELECT to complete. Then until the SELECT completes, the replication will be completely blocked, potentially causing massive slave lag.