_ios_fill_fake_data sometimes generates points where the speed is greater than the speed of sound

[shankari]

In test TestPipelineRealData.testIosJumpsAndUntrackedSquishing, for the second section, we insert an entry because there is a large iOS gap

2023-01-24 21:54:17,644:DEBUG:4623691264:Found 1 large gaps, filling them all
2023-01-24 21:54:17,644:DEBUG:4623691264:Found large gap ending at 8, filling it
2023-01-24 21:54:17,644:DEBUG:4623691264:start = 7, end = 8, generating entries between 1469075385.999996 and 1469075445.999997

And this apparently inserts a point which is over the speed of sound

2023-01-24 21:54:17,737:INFO:4623691264:After first round, still have outliers    index  ...         speed
1      1  ...  6.616681e+00
9      8  ...  3.411318e+07
After first round, recomputed max = 34113180.058238134, recomputed threshold = 1.7570292255633762

This is because the timedelta for the point is very small

Distance delta = 51.28756532378865 and time delta = 41.999999046325684
Distance delta = 22.49458343565475 and time delta = 60.0
Distance delta = 32.436446584498526 and time delta = 9.5367431640625e-07
Distance delta = 50.28642549839274 and time delta = 40.995925188064575

Although the point is not an outlier

Before fill	After fill

[shankari]

Now that our default data collection speed is every meter, we are less likely to run into this. So deferring an investigation and a fix until later.