Nctoolkit

This is a minor release that modifies version pinning for a couple of dependencies.

This is a major release that introduces some breaking changes.

The major improvement in this release will be to vertical methods. All vertical methods should now work with files with vertical axes that are either consistent or vary spatially. Before some methods only worked with z-levels, i.e. files with fixed vertical levels.

This change will result in a requirement that `vertical_mean`, `vertical_interp` and` vertical_integration` need users to specify whether the vertical levels are fixed spatially, using the `fixed` arg. As a result, calls to these methods without this argument will throw an error.

This is a minor release that significantly simplifies basic arithmetic and logical operations.

Simple methods such as `+`, `-` etc. can now use standard python syntax.

For example, if you wanted to add 2 to a dataset you can now do the following:

`ds.add(2)`

as this instead

`ds+2`

The same goes for logical operators. You can do the following to identify if the values in a datset are below 2:

`ds<2`

whereas you previously had to do this:

`ds.compare("<2")

Note: because nctoolkit methods only modify datasets and do not return datasets, the following will not work:

`ds1+ds2+2`

Instead, you would need to do:

`ds1+ds2`
`ds1+2`

This release smoothes out issues caused by xarray being unable to read files. This caused some issues in methods where xarray is used as a backend. `open_data` will no longer through an error in when xarray cannot open the file (xarray is used to parse times), and `check` will provide a meaningful check.

No other changes have been made in this version and it is fully compatible with the previous release.

Fast release that pins the matplotlib version. Fixes an issue that meant an nctoolkit dependency did not import, which broke nctoolkit.

Also features a couple of under-the-hood improvements to methods.

This is a quick release that fixes an issue (caused by a dependency update) that results in multi-variable datasets not plotting on occasion.