Displaying sum memory usage from all workers when parallel runner is used

[Wirone]

In #7777 it was omitted on purpose, to not make PR even bigger and harder to grasp. But to display proper info about memory usage we need to pass data about peak memory usage from workers when they're closing and sum it in the main process. Currently parallel runner displays memory usage only from main process, which may be misleading (but is not critical though).

[keradus]

was omitted on purpose

next time, please raise it explicitly as known TODO, rather than waiting for kerad to notice it ;)