Wrong measurement of query runtime

[yihozhang]

The correct measurement of query runtime is correct for single-thread, but wrong for multi-threads:

• With intra-query parallelism, it underestimates the time because it only counts the runtime of the main thread, while does not include "runaway" children threads
• With inter-query parallelism, it overestimated the time because it counts the sum of time for each semi-naive variant of the query, while these variants are run in parallel.