Create option to write spread files for serial writers

For file types such as .stl files, .ply files, .obj files and .geo files, we currently move all of the data to process 0, and write out the data files from there. For very large runs, we are running out of memory on node 0 using this technique.

Please add the optional ability, default off, to allow users to spread these files. Since these visualizations may be running on dozens to thousands of cores, give the ability to agregate a node's data from all cores/processes to process 0 ON THAT NODE, or just write out each processes data individually.

Example:

• Linux, master, remote server.
• g1s1-10fps.e.16.[0-15]. All vars on. Apply.
• Filters/ Extract Surface. Apply.
• File/ Save Data. Choose one of the file types listed.

Please work on this as soon as 5.6 is out the door.