xyst

Add Dirichlet BC masks

This changes how Dirichlet BCs are expected from the user. Instead of
listing the side set ids, each scalar component must now be listed for
each side set.  This gives more control over the the individual
component and collects all scalar component under compflow. Example:

  problem slot_cyl
  compflow
    depvar u
    material
      gamma 1.66666666666667 end
    end
    bc_dirichlet
      sideset 1 1 1 1 1 1 1 end
      sideset 2 1 1 1 1 1 1 end
      sideset 3 1 1 1 1 1 1 end
      sideset 4 1 1 1 1 1 0 end
      sideset 5 1 1 1 1 1 1 end
      sideset 6 1 1 1 1 1 0 end
    end
  end

This prescribes Dirichlet BCs for sidesets 1-6 and specifies a mask (0
or 1) for each scalar components whether to set Dirichlet BCs or not.
This prescribes Dirichlet BCs for all flow components for sidesets 1-6
and sets scalar Dirichlet BCs for sidesets 1,2,3, and 5. The actual BC
value is evaluated internally, as defined by the problem, in this case
slot_cyl.

For user-defined problems and time-dependent Dirichlet BCs the mask will
still be useful, to prescribe Dirichlet BCs component-wise, but the BC
value must be defined in some other way.

Other, related, changes:

* Remove the disjoint check from the parser's grammar.

* Sideset ids are now stored as ints instead of strings.

* Remove some unused problem types.

* Tested most of this code with a street canyon case which includes
  dispersion of a transported scalar from a point source and the slotted
  cylinder case.

* Renamed BC type tags to more readable ones.

* The solver now does not interrogate inputdeck< tag::param,
  tag::transport >, which will be eliminated in the future. The number
  of scalar components solved for is set based on the problem
  configuration in the grammar.

* Add transported scalar to history output.

* Add a "physics" scalar source for problem type POINT_SRC, besides the
  existing "math" source term.

* For local time stepping, the converge criterion now checks if the
  residual is larger than 0.0. This is needed to prescribe convergence
  criterion for a scalar that has not yet been released.

* Field output now includes the analytic solution (if defined for a
  problem).

* Problem source terms now return a vector and this can be of variable
  size, since some will define flow variables only, some will also
  prescribe scalars.  Calling code will have deal with this.

* Add slot_cyl test problem.