BUG:Information was being copied incorrectly by/from sub grids relative to the actual output information. This fixes it.

Change the way time is collected and used.

Mike Jackson   Senior Research Engineer
Innovative Management & Technology Services

This commit was manufactured during conversion from CVS to merge the
end of the ParaView 3.2 release branch.

wrappings are generated. However, note that installing paraview will not install
vtkpython execultable. Also all the python extension modules will be installed
at the same location as the libraries and not use the python disutils as it does
when one installs from VTK directly.

If re-parse is required, delete Internals beforehand.

ENH:Added a SetTimeValue method, for supporting multiple time steps in the same xdmf file.
Added lag AppendGridsToDomain, to allow multiple grids into one file.
Added a flag InputsArePieces, this allows blocks(pieces)  to be added one by one and then all written as one large dataset.
Added FullGridSize  method to support piece writing.
Various other small bug tweaks addressed with extents and hdf5 read/write issues

BUG: Removing the origin setting was making the individual blocks appear in the wrong place and causing the image tests using multi-block to fail. Might have a problem with parallel load, but this will fix tests for now.

BUG:vtksys::ConvertToOutputPath was adding quotation marks around file paths that have spaces. This broke some tests for xdmf Iron in pv3. Still need to fix a couple more tests.

BUG: When merging changes prior to committing, I missed one origin setting fix. Removed a stray origin set to get parallel pieces appearing right

ENH: Add Time support to Reader. We check for CollectionType attribute on the main top level grid, and if it is temporal, we collect the list of time values from each sub grid. This is exported. When Time values are requested we output the correct sub grid.
Lots of internal changes to tidy up the code and handle information.
BUG: Several bug fixes for reading in parallel. Extents were not always correct. Seem to be ok for image data now

BUG: If we have compiled with Parallel IO and MPI, but are actually runnin a serial only process, the Parallel file access needs to be disabled.
Add a check in the XdmfHDF constructor to set UseSerialFile if either MPI is not initialized, or nprocs<=1