case-lib: refactor kernel logs collecting

[ekurdybx]

Refactor kernel messages collecting:

• save journalctl logs from the start of the test, to avoid losing them in unexpected scenarios
• move setup_kernel_check_point function to the start_test func, instead of calling it in every test in different place
• note: some tests setup more kernel checkpoints than one, and check logs for errors every iteration etc. These setup_kernel_check_point calls remain unchanged - the general dmesg.txt since the first checkpoint still will be available though
• in case of kernel checkpoint not being set, we collect all the journalctl logs
• additional: fix shellcheck issues

[SzymonRichert]

LGTM

[majunkier]

There were already disscussion that pkill -9 should be solved different way. I know there are other files containing this solution but we could do it other way. Like :
Store PID when starting it:

journalctl_cmd ... -f >> dmesg.txt &
PID=$!

and stop only this

kill -TERM "$PID"
wait "$PID"

This avoid killing unrelated journalctl sessions from other jobs

[majunkier]

Moved checkpoint setup and kmsg collector to run after the EXIT SIGTERM trap. This ensures collector lifecycle is owned by main test flow and always cleaned up by exit handler.

   setup_kernel_check_point
    func_kmsg_collect

    if [ "$SOF_TEST_PIPEWIRE" == true ]; then
        func_lib_enable_pipewire
    fi