Imagedata

Changed
* Study class: studyDate/studyTime are now datetime.datetime instances.
* Patient class: patientSize/patientWeight are now float numbers.
Fixed
* Corrected UID handling from v2.0.0-dev1.

Added
* Patient and Cohort classes: Patient is a collection of Study instances, while
Cohort is a collection of Patient instances.
* The dicom read has been modified to keep the UIDs of the input files.
This way SOPInstanceUIDs are correct when you later want to look up SOPInstanceUID from a DICOM PR.
* The series write() method now has an option to keep the UID when writing. The UIDs used to be modified at output to indicate that the data might have been modified. Do something like:

a.write(destination, opts={'keep_uid': True})

Added
* Study class: a collection of Series instances.
Sort images into separate Series depending on SeriesInstanceUID.
(https://github.com/erling6232/imagedata/issues/22)
* Simpler import statements in user code: from imagedata import Series, Study
* Add support Python 3.11 (https://github.com/erling6232/imagedata/issues/21)
Changed
* `input_order='auto'`: Auto-detect the sorting of Series,
depending on which DICOM attribute varies.
The input_orders time/b/fa/te are attempted in order.
* `auto` is now the default input_order.
* DWI images will typically have varying time. Let `b` values override the time stamps
during auto-detect sorting of Series.

Fixed
* Viewer: self.callback_quit was uninitialized when show() was used.
* Series.__array_finalize__(): Copy Header object from any array object which has a Header object.
* Series.__getitem__(): Finalize ret object for all Series instances, independent on slicing status.
* Header.__init__(): No need to set self.axes explicitly.
* Series.axes(): Define RGB axis when last dimension is size 3 or 4 and uint8.
* Series.color: Color property is determined by the presence of an RGB axis. Do not set color property.
* Series.photometricInterpretation: Do not set color property.
* Remove all references to header.color. Color status is determined by the presence of an RGB axis.

Added
* show() and get_roi() works in Jupyter notebook, too.
* In notebook: Series.get_roi_mask() must be called after get_roi() to get the actual mask.

Fixed
* MyPolygonSelector.__init__(): define self._polygon_complete.
* Viewer.grid_from_roi(): Set input_order='none' for resulting mask when follow is False.
* Series.tags: Ensure that tag lists are always Numpy arrays.
* Series: Handle properly the situation where a template has wrong number of slices or tags.