Nexusformat

* Traps `h5py` exceptions when an HDF5 file contains invalid attribute values. This ensures that the HDF5 file will still load.
* Improves writing of generic (*i.e.*, non-NeXus) HDF5 files by removing unneeded class attributes.
* Adds a Jupyter Notebook tutorial to the distribution.

* Fixes a bug when transposing axes in data projections, caused by the restrictions on changing the shape of an existing group item.
* Fixes a bug when renaming NXroot objects
* Fixes a bug in reading axes defined by byte strings
* Ensures that plotting axes that are also NXlinks do not use the parents' name, since these may be generic.
* Makes deep copies of linked items use the children's name, for similar reasons to the previous issue.
* New `sum` and `average` functions are added to NXfields.

* Fixes a bug with renaming fields or groups when the data have not been saved to a file.
* Improves error message when referencing missing attributes in an object without a parent group.

* Improves the handling of external links, especially if an externally linked group contains internal links.
* Improves error messages when overwriting existing fields, which is allowed if the new value has the same shape. The `replace` function must be used when the shapes do not match.
* Allows all the ways of identifying field uncertainties supported by the NeXus standard.
* Uses "." to denote empty axes in conformance with the NeXus standard.
* Fixes the parsing of `axes` attributes that generated a FutureWarning.
* Limits the number of recursions used in the `short_tree` property to two.
* Ensures that natural sorting is used in printed trees, *i.e.*, `scan_10` comes after `scan_2`.
* Fixes a bug in the `NXstack` script, which could cause missing frames at the end of the stack.

* Catch errors reading in NXfield values that prevent the loading of a NeXus file from completing.
* Fixes a bug with plotting data with linked fields.
* Fixes a bug causing infinite recursions when accessing link attributes.
* Adds an NXfield replace function to force a change in the size and/or shape of data already saved in a NeXus file.

* An attribute that is written as a size 1 array is now converted to a scalar. This fixes the reading of size 1 string arrays.