sandbox: lazy-load Project data

A Project can be pretty large on disk - up to several megabytes of data in a single file.

This can make all sorts of related operations slow.

A great idea from @doc-han is to provide some sort of streaming of data from a project file. Because, for example, openfn project only needs to top metadata of a project - it doesn't need to load everything into memory.

I'd suggest we implement a mixture of static and lazy properties on a project:

• All basic project metadata (name, timestamps, parent) is always available
• The list of workflows is always available
• The steps of a workflow should be lazy loaded
• The expression of a step should be lazy loaded

What this basically means is that some functions of the Project and Workflow class should become async, and should only load data when it's actually needed. For example, when doing a partial merge of two projects, I only need to load the relevant workflows - not all of them.

I guess basically anything that gets checked out onto disk should be made lazy. Or should it just be step expressions? Thinking about it, it's probably only the expressions that need to be lazy. The actual workflow structure should be small in most cases?

Start by locating the Project and Workflow classes and the openfn project entry point. Review how project files are read and how partial merges access workflows. Done means basic metadata and the workflow list remain available without loading all step data, while required step expressions load only when needed.