Dataframe view handling of timeline is broken

[abey79]

The timeline view appears to incorrectly behave when the timeline it's configured with does not exist. Unfortunately, that's a very common situation:

1. run some logging script with a custom timeline
2. configure a dataframe view set on that custom timeline
3. run the logging script again
4. 💥

In release build: 💥 == the dataframe view timeline switched to log_time
In debug build: 💥 == panic from a debug assert

What seem to happen is that when the 2nd recording is created but before the first data reaches the viewer, the dataframe view doesn't see the timeline it's configured for and incorrectly handle the situation.

This can be reproduced with the simplest of the script:

import rerun as rr


def main():
    rr.init("rerun_example_bug_dataframe")
    rr.connect_tcp()

    rr.set_time_sequence("another_timeline", 0)

    rr.log("pt", rr.Points3D([0, 0, 0]))



if __name__ == "__main__":
    main()