Dessia-common

Changed

- Reference path is now given all the way down to plot_data
- DessiaObject kwargs in init are deprecated

Fixed

- serialization (to_dict) use some uuids instead of paths for references.
- Workflow : Serialize Imposed Variable Values
- Workflow : to_script Export now write only class name instead of full_classname
- Hot fix for is_valid method of workflow's Block
- Core : add a boolean for platform checking in check_list. Fixes a problem with testing classes not existing
- Reference path for datatools Dataset and ClusteredDataset

Refactor

- Move ./utils/serialization to ./serialization to avoid cyclic-imports

CI

- add a check to enforce update of changelog in PR
- code_pydocstyle.py checks daily instead of weekly
- Add a time decrease effect for pylint

Performance

- Conform doc for many parts of dessia_common
- 100% coverage for clusters module
- cache inspect.fullargspecs calls
- Add trivial checks for simple types
- Avoid redundant serialization in display blocks

Build

- Update python minimum version to 3.8 (3.7 was not supported in fact)
- Update scikit learn to minimum 1.2.0 for the parameter metric in Clustering

Tests

- Add backend backward import tests in order to warn when imports are changed

Chores
- Merge back master to dev
- Docs weekly decrease
- Fixed all remaining pydocstyle errors
- Docs new rules
- More docs

Fixed

- Workflow pipe order when copy
- Diff Dict is now robust to uneven arguments commutation
- Fix path deepth when dict misses keys

Changed
- Refactor copy_pipes and nbv checking

Performance

- Cache for fullargspec and deserialize sequence as comprehension list

- non serializable attributes were not working
- wrong import to sklearn -> scikit-learn

Fixed
- jsonschema bug
- time rendering on workflow

Added
- FiltersList class
- Easy access to Dataset with getitem, len, add, extend, sort
- filtering method in Dataset that calls a FiltersList
- Documentation for Dataset (previously HeterogeneousList), ClusteredDataset (previously CategorizedList), DessiaFilter, FiltersList
- pareto front and parallel plot for Dataset
- Sort parallel plot axis with correlation coefficient (experimental algorithm that seems to work)
- Metrics for Dataset (previously HeterogeneousList) and ClusteredDataset (previously CategorizedList)
- Centroids for ClusteredDataset (previously CategorizedList)
- Nearly all required tests for all these developments

Breaking Changes
- Change attribute "operator" of DessiaFilter to "comparison_operator"
- Change name of file "cluster.py" to "datatools.py"
- Move Dataset in "datatools.py"

Changes
- Add 'logical_operator="and"' attribute to workflow.block.Filter
- Improve workflow._data_eq

Performance
- DessiaFilters and FiltersList: A priori optimized access to elements so it is really faster than before

Fixed
- Excel Export now used proper length of the cell value
- Fix workflow.copy() issue where a nbv with several pipes became several nbv with one pipe

Added
- Clustering classes

Fixed
- Implement to_script method for workflow class
- Prevent foreach name from being it's iter_input name
- Temporarly remove workflow-state from workflow run display settings