Enable the creation of partial metadata and subsequent updating

[andrewphilipsmith]

The metadata models contain many fields. Of these some will be known early in the pipeline, whilst others cannot be determined until later. For example:

• source_metric_id will be known early in the pipeline.
• metric_parquet_file_url cannot be determined until late in the publishing pipeline.

Therefore we need to ensure it is possible to:

• Create an incomplete metadata object (ie that is one where some fields are None) to store known metadata as it progresses in the pipeline.
• Validitate the metadata object at the end of the pipeline, to ensure that the fields are correctly typed, non-none, etc.