Makes hdf5 and libxml2 dependencies optional.

Apply 17 suggestion(s) to 6 file(s)

contains fix: iteration output was overwritten at each timestep because name
did not include timestep or time.
implementation note: Replaced time by iteration because of missing functionality
in BaseLib::constructFormattedFileName to handle iterations

- rename to makeoutput

The previous implementation would miss a fixed output time
if the time step was repeated.

Since for one process output on multiple
(sub)domains / meshes can be required it is
necessary to manage multiple pvd files. All pvd
files for all processes are stored in a multimap.
The findPVDFile method is extended to find the
correct pvd file for the corresponding domain and
process.

For processes containing multiple subprocesses
(staggered scheme and component transport) it
isn't allowed to contain the process id in the
output prefix template.

The integration point writer are defined only on
the bulk mesh of a process. If other meshes (observation,
boundary, etc.) are given for output, the integration
point data must be output only on the main mesh of
a process.

The fmt library happily accepts std::string without
any conversions.

Dmitri Naumov authored 6 years ago and Dmitri Naumov committed 5 years ago

Include spdlog via BaseLib/Logging.h, which
provides macro replacements for
ERR/DBUG/WARN/INFO...

To access the solution values a dof tables are passed
too.

Fixes multiple copyright messages in the doxygen docu.

ag '\\file' -L --ignore=ThirdParty *Lib |xargs  sed -i 's/\(.*\)\\copyright/\1\\file\n\1\\copyright/'

Used for the time step, iteration, etc.

Only if the output meshes list is empty, write it.
Extract OutputFile structure with path and name.
Add lambda for bulk mesh output to avoid duplication.