Mpfa contributed by DennisGlaeser

This patch reduces dependencies and makes usage of linear solvers more flexible, i.e. for two problems with different typetag

* Linear solvers don't need type tag anymore, in fact they don't need any template params
* the preconditioner block size is now provided for the solve method (see e.g. staggered newton controller)
* the amg has a convencience alias using a TypeTag but is "independent" through indirection via the traits

In case of global caching, the grid variables have to be updated
as otherwise the old solution is written to disk.

* [test][richards] Fix solver include
  (Used to compile before due to indirect include over newtoncontroller.

* make output more clear

* calculate the mass and volume fluxes over two different planes

Use the `DUNE_VERSION*` macros to distinguish between Dune >= 2.6
and < 2.6. Employ the correspondingly correct types, constructors
and function calls for Mappers and ReferenceElements.

The point source computation in FVProblem uses the problem implementation to compute the point source map.
However it is executed in the constructor where the problem implemenation is not fully instatiated yet as it
derives from FVProblem. This patch disables point source computation by default. If you want to specify point
sources you have to manually call problem->computePointSourceMap() now.

Replace all calls of `*_PARAM*` macros by correspnding calls to
`getParam`. Remove the corresponding properties. If `getParam`
is called more than once for one parameter, specify a default value
in [`properties.hh`](dumux/porousmediumflow/sequential/properties.hh).
Otherwise, specify the default value in the call to `getParam`.

* test did not compile for box

* type Fluid does not exists anymore

Gabi Seitz authored 7 years ago and Dennis Gläser committed 7 years ago

implementation is not general and is only used to reproduce a
reference from literature.

* remove test for 1p fluidsystem

* use LiquidPhase directly

* use LiquidPhase directly