Scipion-em-flexutils

- Simplified annotate space viewer for inspecting landscapes in a fast manner
- UMAP now uses ParametricUMAP (needs installation of Tensorflow in flexutils env)
- Flexibility Hub data model moved to scipion-em

- Annotate space becomes the only interactive tool for visualization
- Zernike3Deep with model constrains
- HetSIREN new functionalities (TV, MSE...)
- FlexConsensus fixes
- Tensorflow 2.15.x compatibility

- Zernike3Deep method: Semi-classical neural network for estimation of conformational states based on the Zernike3D basis
- HetSIREN method: Neural network heterogeneous reconstruction for real space
- HomoSIREN method: Neural network homogeneous reconstruction for real space
- DeepPose method: Neural network for angular alignment refinement
- DeepNMA method: Semi-classical neural network for estimation of conformational states based on the NMA basis
- FlexConsensus method: Consensus neural network for conformational landscapes
- DeepElastic method: Dimensionality reduction based on elastic cluster analysis
- Dimensionality reduction and landscape visualization splitted into two different protocols
- Strain and rotation analysis for deformation fields (currently for Zernike3D and NMA)
- Clustering at the structural model space (Zernike3D)
- Optimal number of clusters analysis
Developers:
- New data model for Flexibility Hub including several Scipion dynamic objects
- Flexibility Hub protocols.conf added
- Some protocols might rely on new xmipp-metadata Pypi package to read Xmipp metadata without the need to rely on Scipion/Xmipp
- Tensorflow related protocols will rely on the new tensorflow-toolkit (Flexutils) Pypi package to properly setup the Tensorflow GPU environment