Set Function's local ID on creation from remote schema

[alexpearce]

Uploading a Pipeline looks like:

1. Upload the Function objects.
2. Replace the local_id of each Function object with the remote ID.
3. Upload the some other objects (Model and Variable).
4. Create schemas from the remaining objects (GraphNode and output Variable)
5. Create and push a Pipeline schema.

I think there is a bug when resurrecting a Pipeline. The graph nodes will contain the remote Function IDs thanks to step 2 above, however the Function objects resurrected from the schemas will not have the remote ID as they are uploaded before the replacement. This causes the function search to fail because _node.function is a remote ID whereas _func.local_id is a local one.

This PR is one possible fix: when resurrecting a Function from a schema, override the deserialised local_id with the (remote!) ID included in the schema.