qa: Fix test_with_health_warn_with_2_active_MDSs

qa: Fix test_with_health_warn_with_2_active_MDSs

The test intended to validate the failure of 'mds fail'
cmd on any active mds when one of them has warning.

The commit 221700273a82658c642a282c5761c0cbb00ec5b6
(PR 61554) changes this behavior and allows 'mds fail'
on mds without the warning. The test should have always
failed with this commit. But the test never failed until
tested extensively because the test mostly generated
warnings for both active mdses. Occasionaly, the test
generated a warning on single mds and failed. So it's a
race. This patch fixes the same by changing the following.

 a. Changed the mds_cache_memory_limit to '50K' from '1K'
    as '1K' was to less and generating warning on both the mdses.
 b. Create a directory and pin it a single mds and open 400 files
    in the backend to create cache pressure on one mds.

Also, there are two tests with the same name as
'test_with_health_warn_with_2_active_MDSs' but in different classes
though. So changed the test name to
'test_with_health_warn_on_1_mds_with_2_active_MDSs' to avoid
confusion and indicate what the test actually does.

Fixes: https://tracker.ceph.com/issues/71915
Signed-off-by: Kotresh HR <khiremat@redhat.com>