feature request: don't require node ids provided separately when using projections

[sarmbruster]

When running any algo on a projection we have to supply

1. a list of node ids
2. a list of source id, end id and weight for the relationships

The need for providing node ids separately seems to be redundant. They are already given by source and end ids.

This is esp. helpful if your projection is rather complex so need to run the same projection statement twice.