What to do when VectorOfVectors is too big

[iguinn]

To handle VoV in the processing chain, we use a 2D array to act as a buffer (effectively an AoESA). We NaN/zero pad the end of this array, and we calculate the vector length as a separate variable, which gets used to set the VoV lengths. The question is what to do if the vector length is greater than the second array dimension. Right now we don't have a standards solution, but there are some options:

1. Raise an exception. This will cause a file to fail, which I don't think we want
2. Have processors stop counting once they hit the array size. This will make it un-obvious that something has gone amiss
3. Have processors keep counting but stop filling the AoESA buffers (this is the current solution for bi_level_zero_crossing_time_points). In this case we need to choose how to handle copying into the VoV. One option is to copy NaNs into the array (although for int types it would have to be 0s or 0xDEADBEEFs or something else much less obviously wrong than NaN). Another option is to copy only up to the array size and throw out a warning into the log.
4. Maybe someone has a better idea?

[gipert]

https://awkward-array.org/doc/main/user-guide/how-to-use-in-numba-intro.html?

[iguinn]

This is solved in this PR: legend-exp/legend-pydataobj#109

This makes it so that if the vector length is longer than the AOESA, it will fill the VoV with the full length and pad the missing values with nan for floats, min integer for signed int, and max integer for unsigned int