Refactor and generalize unit operations, particle modeling

[jbreue16]

This meta-issues tracks progress related to the refactoring of chromatography units, including #414 and open ends beyond

TODO

• Deprecate collocation variant of DGSEM particle discretization #411
• Modularize particle DG discretization #406
• Refactor particle operator to be particle type specific #423
• Add binding and reactions to particle operator #424
• Modularize particle discretization in particle operator #431
• Modularize particle model #432
• Add homogeneous particle model class, integrate into LRMP_DG unit #434
• Move film diffusion to particle model #435
• Add generalized 1D column unit #436
• Refactor 2DLRMP unit into generalized 2D column unit #458
• performance benchmarks with old models: 10 runs, required compute times are in the same range but vary from run to run, sometimes old and sometimes new units are better. Same on average.
• Remove obsolete unit operations #464, but allow old unit_type specification (which will still require the new particle interface except particle_type)
• Update to generalized interface #471 - keep the optimized arrow-head FV units, but update interface
• Add FV particle diffusion operator #465
• Decision: we will keep a separate "without pores" unit operation for performance and code readability reasons
  • refactor the interface, change modelbuilder so that this internally separate unit is called
• add bulk FV to generalized units
• make forum post when this is merged on master to announce these breaking changes and explain our release strategy. team post can serve as template/draft. hint to docker images. emphasize that the documentation on master is now breaking all <v6 releases
• finalize documentation
  • get review
  • what to do with the possibility to specify UNIT_TYPE with the legacy names
  • rework modeling section and in that context what to do with paper and cadet-equations?

Follow up issues

• modularize the FV and DG convection dispersion operator (1D), use in generalized unit. That'd result in a FV discretization without arrow head. Note: Move Jacobian recomputation into residualimpl of the modules: remove diagonal entries reset in DG particleDiffOp, change DG Jacobian to addition, move bulk jac to convDispOp. Also add bulk FV to the LRM unit
• Add a CSTR with the new particle interface, see Implement tank with particles with pore diffusion #34
• implement new interface in cadet-process, cadet-verification, cadet-equations