Is there any best practice for dividing a worklight project into streams and components?
A IBM Worklight project should use components for each coherent set of files, like a worklight project (with its containers, html files, css, ...)
Anything generated, like an android project generated to be run and simulated, shouldn't be in a component.
You can start simple with a single stream (which groups all components you need to work), or with one stream per component (but that approach doesn't scale well past 10-20 components, since the streams have no hierarchy)