One of the popular diagnostic tools are slow query logs

Currently the production of these logs is controlled by two parameters: slowms and a profile filter.

And the logic is not obvious: the query is logged after it has been completed if it took more than slowms milliseconds OR it matches the profile filter.

In SERVER-17942 we are going to add slow in-progress query logs that adds even more logic to this.

We need to write it in a single place to use as a reference