[bug] latency shifts when you retrieve the node in the trace tree

[mikeldking]

Clipboard-20250326-010955-847.mp4

[anticorrelator]

check that it's not cumulative in one case and not the other

[RogerHYang]

This is due to SQLite's lack of precision at the sub-millisecond level when computing the latency. The trace query here is returning the latency computed by SQLite, but the second query here first fetches the db record, and then computes the latency value in Python. The issue is that these calculations don't always give the same values at the sub-millisecond level. The Python calculation is more accurate.