Function to drop/flag obviously bad data

[serbinsh]

It would be good to remove some of the hard-coded assumptions in the processing functions to allow for flagging/dropping data where we have obviously broken/bad thermister data (0.9-1.0) during data load. This way we can avoid having to deal with those data in the data processing steps and hopefully avoid issues with erroneous data.

This would require flagging at the start then dealing with issues including when there is no usable data during a time window and skipping that tree in the processing steps.

[DavidsonKen]

I have added functionality to the "sanity_check_fine" function so that it now sets all data which are negative to NA, and also flags data with no heat pulse and changes that whole pulse to NA.

We can still also build in a function to flag bad data from the start, but the automated function should cover most cases of bad data