Could we use something better, may be using extra data instead for fields that should be copied back from the workspace in a lower package?

Yeah this is much better and correct, doing this now.

Also would the top level workspace exist as a package of its own or not?

So I did consider doing this, but the top-level workspace usually does not have any purl fields there, only some extra data wrt source/licenses, see https://github.com/nexB/scancode-toolkit/blob/support-cargo-workspaces/tests/packagedcode/data/cargo/cargo-with-workspace/Cargo.toml#L12 for example. So for now we are not adding top-level packages here, but we can revisit this if we get better examples. But if there are other manifests of other flavours, we'll be adding packages for those.

I've removed the improved package assembly part from this PR and pushed this as a separate branch at: https://github.com/nexB/scancode-toolkit/tree/update-package-assembly, since this was not ready to merge. Otherwise this was reviewed and approved, so merging!