🐛 Metadata updates assume that keys are the same between old and new version

[MatMoore]

Describe the bug.

When updating example_prison_data_product_metadata to add a retentionPeriod, I got this stacktrace:

[ERROR] KeyError: 'retentionPeriod'Traceback (most recent call last):
  File "/var/task/update_metadata.py", line 42, in handler
    new_version = version_creator.update_metadata(new_metadata)  File "/var/task/versioning.py", line 98, in update_metadata
    state = metadata_update_type(metadata)  File "/var/task/versioning.py", line 135, in metadata_update_type
    changed_fields = data_product_metadata.changed_fields()  File "/var/task/data_product_metadata.py", line 245, in changed_fields
    if new_version[field] == latest_version[field]:
 | [ERROR] KeyError: 'retentionPeriod' Traceback (most recent call last):
   File "/var/task/update_metadata.py", line 42, in handler
     new_version = version_creator.update_metadata(new_metadata)   File "/var/task/versioning.py", line 98, in update_metadata
     state = metadata_update_type(metadata)   File "/var/task/versioning.py", line 135, in metadata_update_type
     changed_fields = data_product_metadata.changed_fields()   File "/var/task/data_product_metadata.py", line 245, in changed_fields
     if new_version[field] == latest_version[field]:
-- | --

This logic assumes the set of keys is always the same when updating the metadata, but this is not the case, since

1. the metadata schema can evolve
2. the user may add or remove fields that are optional in the schema

To Reproduce

1. Create a data product with minimal required fields
2. Update the data product to add a new field

Expected Behaviour

The metadata should be updated successfully and a new minor version created

Additional context

No response