Bug when decoding: "duration_complete event must have start tick (166428285) greater than end tick (166428276)"

Nobody has claimed this yet.

When running `./magic-trace run -working-directory workdir -full-execution -multi-thread ./dotnet/dotnet -- benchmark/Abc.TradingPerfTests.dll`

magic-trace end up throwing when writing the trace with the error:
`"duration_complete event must have start tick (166428285) greater than end tick (166428276)"`

Thanks for helping us make magic-trace better!

Please consider answering any subset of the following questions, based on what you think
is most relevant to your report:

> - How can we, magic-trace developers, reproduce the issue you're seeing?

If needed I can provide a zip that contains everything to reproduce the issue.

> - What version of perf userspace tools are you running? `perf --version`

`perf version 6.8.12`

- What CPU do you have? `cat /proc/cpuinfo`

`Intel(R) Xeon(R) Gold 6146 CPU`

Start by reproducing the failure with the provided magic-trace command, perf 6.8.12, and the reported Intel Xeon CPU. Trace the decoding and trace-writing path that reports the inverted duration_complete ticks, then verify that the same workload produces a valid trace without throwing.

Written by the indexing model from the issue text.