Changed

Improved the default preconditioners for poroelasticity for improved scalability.
Improved the default preconditioners for elasticity with faults for improved scalability.
Replaced use of Cubit journal files with Cubit Python scripts for several examples, and use the skeleton sizing function to increase cell size with distance from the fault.
Removed ParaView Python scripts (replaced by pylith_viz)

Added

New 2D and 3D crustal strike-slip faults examples based on the 2019 Ridgecrest earthquake.
New 2D subduction zone outer-rise faulting example examining poroelastic response to bending stresses.
New pylith_viz utility for plotting PyLith results.
Updated examples/strikeslip-2d and examples/reverse-2d to demonstrate use of uniform refinement and higher ordder discretizations to improve resolution of solution.
Documentation
- Developer Guide: Added description of the process for adding event logging and evaluating performance, including performance logging.
- Developer Guide: Added checklist for what is needed when contributing examples.

Updated examples/subduction-3d Steps 6, 7, and 8 to work with PyLith v3+.
Fix performance bottleneck when reading Gmsh files.
Remove extra (wrong) kernels for poroelasticity when using state variables.
Update Python unit tests setup for compatibility with the current unittest API (use load_tests()).

The new default preconditioner for simulations using elasticity and faults can cause convergence issues when running in parallel in which fault faces lie on the boundaries between processors. The workaround is to use the previous preconditioner provided in share/settings/solver_fault_fieldsplit.cfg.
The default PETSc options provide a computationally expensive preconditioner when solving incompressible elasticity problems. We expect to have a more optimal preconditioner in the next release.