Nctoolkit

This is a minor release with a couple of method upgrades.

The `subset` method now lets users subset hours. For example `ds.subset(hour=1)` will select the first hour of the day.

The match point methods have also been modified so that the dataframe sent to them does not have to strictly match names, so that names such as "longitude", "Longitude" will automatically be decoded as longitude.

This is a minor release. It introduces a couple of new methods: `rolling_var` and `rolling_stdev` for calculating rolling variances and standard deviations.

All code written for the previous version of nctoolkit will be compatible.

This is a major release that introduces some breaking changes.

All methods that carry out temporal averaging of any sort will now align output times to the right. This applies to methods such as `tmean` and `rolling_mean`. The internals when `align = "left"` option have been modified, as the CDO call was sometimes giving incorrect results.

This is minor release, which features some method enhancements and bug fixes.

It improves the abilities of temporal methods, giving users the ability to select how they want times in output to be aligned. For example, if you are calculating a rolling mean, you might want the output times to be the first, middle or final time in the temporal window. This release
will add that ability to nctoolkit's temporal methods. Previously nctoolkit used CDO's default methods, and did not allow users to do anything else. By default, output dates will be aligned to the middle time step.

The `match_points` methods were throwing an error when there were non-unique vertical values. This is now fixed.


Some improvements have been made to package internals.

A quick bug fix release.

Fixes an issue with imports that caused `match_points` to fail sometimes. Also, fixes `subset` so that it will allow * and ? wildcards for variables.

This release improves error handling when nctoolkit fails due to an error in Climate Data Operators. Datasets will now be reset to the starting point when this happens and a clear message will given to users.