Change where min_years cutoff is applied

This MR moves where a minimum years threshold to rank streamflow is applied from the calculate percentile threshold function to the categorize flows function. Per #74 (closed), this makes things a little cleaner as the functions to calculate percentile thresholds now test how much data is available before calculate percentile thresholds. So, a uniform cutoff of number of years a site has data for in order to rank streamflows makes more sense where we categorize flows.

MR also fixes bug on NAs when interpolating percentiles for streamflow values identified in #78 (closed)

This MR should close #74 (closed) and #78 (closed)