Hand FIM improvements

[groutr]

The biggest change is selective column loading for csv files. This could potentially significantly reduce memory usage.

The rest of the changes are relatively minor.

• np.where(condition) is an inefficient way of masking an array. It's about 2x slower and allocates an extra array.
• sorting values is not an efficient way to find the max.
  ...and other minor fixes that simplify the code.

[RobHanna-NOAA]

We might want to double check which fields from the that table show up in the UI. There may be UI columns that the field want to see. Update here on what you find as I can take it back to FIM-Dev and find out why they are submitting those columns to HV.

[groutr]

I'm not sure I understand what you mean. The code is not discarding any columns that were not already being discarded.

[RobHanna-NOAA]

doh.. on nothing. I am a dork. I was mis-reading the FIM-dev input dataset. :)

[shawncrawley]

Looks good - thank you!