Niondata

-------------------
- Improve compatibility with HDF5 backed data (use numpy.copy instead of d.copy).

-------------------
- Enable support for Python 3.10.
- Fix issue for cross correlation using only first image.
- Fix regression with template register using rounded position.
- Extend sequence trim/integrate to work on sequences of collections.
- Fix half-pixel offset in register_template.
- Allow a mask in register_template for fine-tuning maximum finding.

-------------------
- Enable strict typing.
- Drop support for Python 3.7.
- Add auto-thresholding functions.

--------------------
- Optimize element data for sequence + collection case.
- Change sub-pixel registration method to parabola fit to improve speed.

--------------------
- Add affine transform function and optional order parameter to warp.
- Allow zero-dimensioned (scalar) data and metadata objects.

--------------------
- Make special case of C(1) D(1) < 16 from element data function optional.