Pytme

Version 0.2.0 provides a new interface for dealing with orientations in 3D space, quality of life improvements to the GUI and fixes a critical performance issue introduced in a previous version.

**Enhancements**
- New Orientations class provides a convenient interface for converting between different file formats, namely tsv, relion star and dynamo tbl.
- Severe runtime decrease (2-3 times) for scoring methods affected by an issue in cupy.std, including scores computed using tme.matching_exhaustive.corr_scoring and tme.matching_exhaustive.flc_scoring.
- Revised interface for various command line options. New sampling and filtering functionalities in match_template.py
- The preprocessing GUI can now automatically import pickle files created by match_template.py, read all formats supported by tme.density.Density.from_file and provides additional functionalities for alignment.

**API Changes**
- Moved Orientations class from postprocess.py to tme.orientations.Orientations.
- tme.matching_optimization was refactored to allow individual scoring objects to be setup more easily and to facilitate the usage of different optimizers.
- tme.preprocessing.Compose introduces composable filters that were made accessible via match_template.py.

**Documentation**
- init methods of classes defined in tme.matching_optimization are correctly displayed.
- Inheritance structure is made more obvious.
- Attributes are now correctly documented for dataclasses.

v0.1.9 fixes a bug introduced in v0.1.8 that impedes processing large non-floating point data MRC files.

Version 0.1.8 provides a series of enhancements over the 0.1.7 release.

Enhancements:

- tme.matching_data.subset_array makes use of file metadata for contrast inversion. Previously, the relevant statistics were computed on subtomograms directly, which resulted in patchy score spaces if the local intensities deviated strongly from the global intensities.

- tme.matching_data.subset_array makes use of a different padding strategy that reduces the influence of boundary effects.

- The scores "FLC" and "FLCSphericalMask" no longer internally convert template masks to binary, but rather use the floating point values when normalizing the template.

- Tilt angle based wedge masks are computed more efficiently.

- The preprocessor_gui now supports low/highpass filtering based on resolution.

- A Dockerfile for containerized deployment on GPU hardware was added.

- The integration chapter was moved from the postprocessing section to a new dedicated section.

Bugfixes:

- Fixed and issue where the center of tube masks would not be propagated correctly

Version 0.1.7 fixes bugs discovered in the 0.1.6 release.

Bugfixes:
- tme.density.Density.from_file factors in extended headers in CCP4/MRC files when reading subsets from disk. This might have caused issues for tomograms reconstructed using AreTomo, if the tomogram was split into subsets during template matching.
- match_template.py correctly reports target mask in the output logs.
- tme.matching_exhaustive.scan_subsets produces warnings if the template is larger than the target and padding is deactivated, which can cause shifts in generated scores spaces.

API Changes:
- tme.density.Density.centered now performs linear interpolation when moving the center of mass to the center of the data attribute, rather than relying on whole integer translations.
- tme.preprocessor introduces a variety of new filtering methods.
- Edge and fourier padding logic was moved from tme.maching_exhaustive to tme.matching_data
- Introducing experimental MLX backend for apple silcone devices

Documentation:
- Minor wording changes.