Diff for first interpolated sample may be non-zero

The following discussion from !43 (merged) should be addressed:

• @jmfee started a discussion:

  It's possible the first bin actually had rainfall.

  • One approach would be to use the value precip[0], but that would be all the rainfall reported during that bin instead of the minute.
  • Another approach would be to handle this problem outside this method: in process_fire, check for rainfall in the first data sample. if rainfall occurred, request a slightly larger time interval for processing station data and then trim the returned data frame to the originally requested starttime/endtime.

Relates to #52 (closed)