Optimize file reading

[plaplant]

We've been noticing that loading data has been very slow for the data notebooks. I think right now, this is primarily due to using the UVData.__add__() method to combine data instead of the much faster UVData.fast_concat(). For now, the change may be as simple as specifying the axis="blt" keyword argument to the current UVData.read() method calls. This change would use the faster method under the hood, without sacrificing the ability to skip bad files.

In the longer term, we probably also want to implement select-on-read for specific cells/tasks (e.g., when plotting auto spectra, we should only read the autos with the ant_str="auto" keyword). If I'm understanding the code in utils.py correctly, we might be reading in all the data for a single night, and then performing various selections. This will become prohibitively expensive very soon, so we should make changes now.