Fix CSV export for mixed content with Vocabulary metadata

Steps to reproduce:

1. Deploy the jar from https://jira.nuxeo.com/secure/attachmentzip/unzip/152948/100459%5B5%5D/
2. Go to a Workspace
3. Create a document whose type is Doc1 and fill all the custom metadata
4. Create a document whose type is Doc2 and fill all the custom metadata
5. Do a CSV export for this workspace with  the 2 new documents

Observe that the content is not aligned with the column headers for Doc1

The problem comes from the "write" method at https://github.com/nuxeo/nuxeo/blob/805d23810353a23253256fd8c90c7a05128ac832/nuxeo-features/nuxeo-platform-csv-export/src/main/java/org/nuxeo/ecm/platform/csv/export/io/DocumentPropertyCSVWriter.java#L65 which prints a single"null" if the property is null (for example when the document does not have this schema) whereas 2 columns were written to the headers when the property is a vocabulary (one for id, one for label)

Expected behavior:

All content should be aligned with the respective columns.