Backoff for sync preflushes

Currently if a log-writing thread finds that the checkpoint age is in the sync preflush zone, it will attempt to advance the checkpoint itself by issuing a flush list flush batch unless one is running already. After the page cleaner tuning [1] this feature hinders more than helps: the cleaner thread knows that the system is in sync preflush and will perform furious flushing itself. The thread doing its own flushes will only contribute to mutex pressure and use CPU. It is better for the query threads to wait for any required flushes to complete instead.

[1] is a prerequisite, even though the page cleaner would have performed the necessary flusing in its stock version too, is that with [1] fixes the cleaner will not be performing LRU flushing too long, prolonging the needing log space query thread wait.

Implement by adjusting log_preflush_pool_modified_pages() to wait until the oldest unflushed page advances past the target LSN. This is controlled by a new bool option, innodb_foreground_preflush, default on.

[1] https://blueprints.launchpad.net/percona-server/+spec/cleaner-tuning-5.6