This bug happened because the maps of covering keys for mergeable derived
tables/views was not recalculated after the derived tables/vies had been
merged into the main query.

restore the status quo from before the microsecond patch

- Testcase

- Make advance_sj_state() not to attempt building duplicate removal strategies
  when we're doing optimization of an SJM-nest.

they were deleted and re-created.

- in make_join_select(), use the correct condition to check whether the current table is a SJM nest (the previous 
  condition used to be correct before, but then sj-materialization temp table creation was moved to happen before
  make_join_select was called)

This bug is a consequence of the fix in the function add_ref_to_table_cond
for LP bug 826935 that turned out to be not quite correct: it tried to AND
the same generated condition with two different other conditions.
This patch creates a copy of the generated condition if the condition needs
to be ANDed with two different items.

Also, fix code to get physical memory size.

and disable feedback plugin by default, if it's compiled in.