Initialize arrays for speed savings

[dnowacki-usgs]

This approach improves run speed by 50% for a ~10 GB Vector file.

CC @rbales-usgs

[dnowacki-usgs]

On my machine, the run speed went from MIA23TR202d-raw.cdf 527.98s user 273.30s system 94% cpu 14:07.49 total to MIA23TR202d-raw.cdf 319.50s user 38.97s system 88% cpu 6:44.44 total. As far as I can tell, the outputs are identical.

[rbales-usgs]

That's a great addition! Thanks for working on that.

PR #257 I submitted last week was primarily submitted to get the SNR and COR thresholds in. Along with using trim_maxabs_diff on burst shaped data. The removal of the Z dims and changing float64 variables to float32 was mostly triggered by the time it takes to write out a continuous vector .nc file. It doesn't take as long to write out the Burst vector nc files. But, if you rather those changes not happen I can edit that branch so that it only includes the addition of the QAQC features.

I will say, dropping the Z dim from the variables makes the dataset a lot more user friendly (faster calculations and don't need to use squeeze when plotting). But if you think it is important to keep them associated with the variables then we can. One thing Steve suggested was make the Z dim a variable attribute instead of a dimension for a variable.