Bug 2318289

Header read time are now reported accurately Previously, the header read time was calculated based on the throttle timestamp (`throttle_stamp`) instead of the receive timestamp (`recv_stamp`). While the throttled event preceded the header read event in the event queue, the throttle timestamp was older than the receive timestamp. This indicates a logical error in how the header read time is determined when throttling occurs. The header read time must reflect when the header was *received*, not when the associated operation was *throttled*. As a result, incorrect header read times were reported. This led to inaccurate performance metrics, making it difficult to analyze and optimize system performance. It also triggered false alarms in monitoring systems that relied on accurate header read time data. Furthermore, it masked actual performance issues if the reported header read times were artificially inflated by throttling delays. With this fix, the code is modified to use the receive timestamp (`recv_stamp`) as the basis for calculating the header read time, regardless of a throttling event. The header read time now represents the time the header was received and not the time the corresponding operation was delayed. Hence, accurate header read times are reported, reflecting the actual time the header was received. This will provide more reliable performance metrics, improve the accuracy of monitoring systems, and enable better analysis and optimization of system performance.