Update to generalized interface

[jbreue16]

double check and document interface changes

• remove enable_dg flag, eigen will become a required dependency
• implement generalized interface for old FV and LRM_DG units. also change this in createLWE, breakthrough.py, etc
• implement generalized modelbuilder, only new interface should be exposed
• update names of input fields
  • rename: PAR_BOUNDARY_ORDER -> FV_BOUNDARY_ORDER, remove par suffix from particle cell/element spacing types, e.g. EQUIDISTANT_PAR
  • remove field PARTICLE_TYPE. instead, we will have explicit bools for what effects are included in the particle model, i.e. has_film_diffusion has_pore_diffusion, has_surface_diffusion, which default to false. add automatic check for impossible/unsupported combinations. if a field is set to false, the corresponding parameter, if provided, is ignored.
  • rename PAR_DIFFUSION to PORE_DIFFUSION and PAR_SURFDIFFUSION to SURFACE_DIFFUSION
• update field UNIT_TYPE :
  • if GRM, LRMP are specified, the respective particles are used automatically but check if the fields are inconsistent if they exist.
• add documentation, see below for interface changes
• [x] make modified newton the new default, see Change default for USE_MODIFIED_NEWTON to true #478

[schmoelder]

remove field UNIT_TYPE from standard configuration, but keep it for backwards compatibility. if GRM, LRMP are specified, the respective particle should be used automatically. Instead, for the standard configuration we need a new field indicating the dimensionality, and maybe the direction of the flow or, even better, the device geometry/type. what to do with FV units?

I would not recommend removing the UNIT_TYPE field. We might want to add different unit operations (e.g. filtration) which are not part of the generalized column family.

[jbreue16]

I would not recommend removing the UNIT_TYPE field. We might want to add different unit operations (e.g. filtration) which are not part of the generalized column family.

It'll stay, the comment was not updated. Old specifications such as GRM, LRMP will imply the particle type and $N^p=1$, which is not just backwards compatibility but a feature of the interface. The new input for a generic column will be something like AXIAL_FLOW_COLUMN, RADIAL_FLOW_COLUMN, indicating the geometry. Another, separate field will be used for the dimensionality and default to 1D

[jbreue16]

Interface changes:

• UNIT_TYPE can now be one of the legacy names or COLUMN_MODEL_1D, _2D, RADIAL_COLUMN_MODEL_1D and so on; ie different flow directions/geometries are added as prefix, dimensionality as sufffix
• field NPARTYPE defaults to zero, i.e. becomes mandatory when particles are used, except when old unit_types are specified, then default is 1
• new group particle_type_000 for every particle type with the respective index
• NBOUND is moved into the particle group(s)
• multiplexes (e.g. FILM_DIFFUSION_MULTIPLEX) exclude partype multiplex. some multiplexes which only had partype, e.g. ADSORPTION_MODEL_MULTIPLEX, REACTION_MODEL_PARTICLES_MULTIPLEX are deprecated.
• New fields are used to specify parameter dependence as used in parameter sensitivities:, FILM_DIFFUSION_PARTYPE_DEPENDENT, BINDING_PARTYPE_DEPENDENT and REACTION_PARTYPE_DEPENDENT, PAR_SURFDIFFUSIN_DEP_PARTYPE_DEPENDENT, PAR_RADIUS_PARTYPE_DEPENDENT, PAR_CORERADIUS_PARTYPE_DEPENDENT, PAR_POROSITY_PARTYPE_DEPENDENT, PORE_ACCESSIBILITY_PARTYPE_DEPENDENT, PAR_DIFFUSION_PARTYPE_DEPENDENT,
  the internal MultiplexMode remains the same and is derived as before but with the additional information from the new partype dependence field.
• parameter dependencies for sensitivities now must be specified via PAR_RADIUS_PARTYPE_DEPENDENT etc. otherwise, parameter is assumed to be dependent on the particle type.
• REACTION_MODEL_PARTICLES -> REACTIN_MODEL in each particle group
• if needed, particles have their own discretization group (only General rate particle for now)
• initial conditions of $c^p, c^s$ were moved to particles: INIT_CS (formerly INIT_Q) and INIT_CP. in the case of particle type independent bindings, it is checked if initial conditions are the same across particle models
• NPAR -> NCELLS
• NCOL-> NCELLS
• SPATIAL_METHOD is mandatory for both bulk and GRM particle
• multiplexing: only 0 and 1: 0 -> component dependent, 1 -> section and component dependent. Sensitivity dependence is handled via new field {param_name}_PARTYPE_DEPENDENT
• introduced fields HAS_FILM_DIFFUSION HAS_PORE_DIFFUSION HAS_SURFACE_DIFFUSION
• renamed PAR_DIFFUSION to PORE_DIFFUSION and PAR_SURFDIFFUSION to SURFACE_DIFFUSION
• renamed COL_DISPERSION_AXIAL in 2D units
• document choice of linear solvers for generalized units

[jbreue16]

Seems like the CI is triggered for commits changing files listed under paths-ignore on PRs that have changes in other places too