Performance optimization

For very large models there is a tendency of slow processing, in particular in interactive sessions. This is because the persistent model is loaded to memory to be extended.

There are optimization suggestions at several places in the code (workflow_executor.py, fireworks.py). These are mainly to avoid loading Fireworks and Workflow documents.

One can also avoid loading the whole model in the special cases of only print and view statements including only one reference.

Also a timing profile can be generated to identify bottlenecks.