match_and_remove_trend is slow

[jbusecke]

I assume this is due to the fact that is actually needs to compute some of the data for the comparison of missing values, but I wanted to flag this here for potential future performance tuning