Bug#27294009 XCOM ENTERS AN INFINITE LOOP IN CASE OF A REQUEST_RETRY (post-fix)

Problem
=======
gr_start_onboot_fail_restore_read_mode has been consistently failing on
Windows.

Analysis
========
gr_start_onboot_fail_restore_read_mode fails consistently when the same MTR
worker executes gr_start_group_without_boot_node immediately before
gr_start_onboot_fail_restore_read_mode.
I am unsure why this particular test order is problematic, but using a private
error log for gr_start_onboot_fail_restore_read_mode solves the issue.

Solution
========
Use a private error log for gr_start_onboot_fail_restore_read_mode.

Reviewed-by: André Negrão <[email protected]>
Reviewed-by: Nuno Carvalho <[email protected]>
RB: 18752

Author: tvale

rapid/plugin/group_replication/tests/mtr/t/gr_start_onboot_fail_restore_read_mode-master.opt

@@ -0,0 +1,2 @@
+--log_error=$MYSQLTEST_VARDIR/tmp/gr_start_onboot_fail_restore_read_mode.1.err
+--no-console

rapid/plugin/group_replication/tests/mtr/t/gr_start_onboot_fail_restore_read_mode.test

@@ -45,8 +45,7 @@ SET SESSION sql_log_bin= 1;
 --echo #    replication
 --echo ############################################################
 
---let $assert_file= $MYSQLTEST_VARDIR/log/mysqld.1.err
---let $assert_only_after = CURRENT_TEST: group_replication.gr_start_onboot_fail_restore_read_mode
+--let $assert_file= $MYSQLTEST_VARDIR/tmp/gr_start_onboot_fail_restore_read_mode.1.err
 --let $assert_count = 1
 --let $assert_select = Timeout on wait for view after joining group
 --let $assert_text = Found the expected Group Replication start error in the server log