Yet more test changes to prepare for the hypergraph optimizer:

 - Add --sorted_result to several test files.
 - Use the new --skip_if_hypergraph directive to skip some statements
   whose results depend on the query plan chosen or similar.
 - Add a few blank lines between some queries so that diff(1) will
   have an easier time making diff hunks that MTR's hypergraph error
   detection will parse.

The latter case is a bit tricky. Imagine two skipped queries in a row:

  EXECUTE stmt;
  EXECUTE stmt;

Normally, the output would look something like this:

  EXECUTE stmt;
  response for s1
  EXECUTE stmt;
  response for s2

but with --skip_if_hypergraph or similar, it would be instead:

  EXECUTE stmt;
  <ignored hypergraph optimizer error>
  EXECUTE stmt;
  <ignored hypergraph optimizer error>

We'd expect the diff to be then

   EXECUTE stmt;
  -response for s1
  +<ignored hypergraph optimizer error>
   EXECUTE stmt;
  -response for s2
  +<ignored hypergraph optimizer error>

but if we're unlucky, diff(1) could choose to parse this as

   EXECUTE stmt;
  -response for s1
  -EXECUTE stmt;
  -response for s2
  +<ignored hypergraph optimizer error>
  +EXECUTE stmt;
  +<ignored hypergraph optimizer error>

mysqltest's logic would then go ahead and ignore the first five lines of the
diff, but then see the “added” EXECUTE stmt; line, erroneously seeing it as a
non-ignorable diff and thus fail the test instead of skipping it. Adding some
extra air helps the diff in making skippable output.

Change-Id: Ib8edcb13c241299dba545b2def2fe47aac4a2df7