Test failures occurring when running the tests with the non-default 4K
page size for InnoDB. The tests fail due to changes in the "disk sweep
cost model" for reading data from disk introduced in WL#6082. The disk
sweep cost is influenced by the number of blocks used for storing the
table which can be different when using 4K page size compared to 16K
page size.

opt_trace.bugs_no_prot_all:
===========================

Cause: this test fails due to the cost estimate in the optimizer trace
       is slightly different due to change in "file size" when using
       4K blocks instead of 16K blocks.

Fix: This is an optimizer trace test. Fixes this problem by only running
     it with the default 16K block size for InnoDB.

innodb_explain_json_non_select_all, innodb_explain_non_select_none,
innodb_explain_json_non_select_none, nnodb_explain_non_select_all:
==================================================================

Cause: These tests fail due to change in execution plan between 16K
and 4K InnoDB page sizes. For 16 page size the failing query is done
as a table scan while for 4K page size it is done as an index-merge.
Before WL#6082 it would use a table scan in both cases. The reason for
the change to do index-merge for the 4K case is that the size of the
table is smaller when using 4K pages compared to using 16K pages and
thus the cost for doing index merge is now lower.

Fix: We need to have the same query plan for both 4K and 16K pages
(since the test has the same result file). For the same tests when run
with MyISAM the queries are done using index-merge. So instead of
"forcing" the 4K InnoDB test to run as a "table scan", we change the
default/16K InnoDB test to run using index-merge by adding more
records to the table. This increases the cost of doing a full table
scan and makes index-merge cheaper than a table scan in both 16K and
4K test case.