This may unexpectedly broke performance reports and provide wrong results due to measurement in runtime.

• it measures a time between exiting of startTimer(), iter++ and calling of stopTimer(). Critical code path is minimized for this flow.
• on properly configured system its value is almost zero: about 10-11 clock ticks (ns, on Linux x86). By comparing to 1'000'000 clock ticks of 1ms sample measurement - it is a nothing.
• other values are not expected (and not observed during the 5+ years). Other values would just break the numbers in the performance report.
• this value is already ignored (forced to 0) for PERF_STRATEGY_SIMPLE which is default since 2014.

Lets remove this unnecessary step of performance tests startup.