Catch errors when reading serialized schemas also for collections #230
Description
If the format of serialized schema or collections change, previously serialized schemas might not be readable with new dataframely versions (e.g. primary_keys vs. primary_key). For schemas, this is already handled by catching errors in the dataframely.schema.deserialize_schema method if strict is False.
It would be nice if this could also be done in collections (dataframely.collection.collection._deserialize_types).
Additionally, in both cases, when these methods are called transitively via a scan_parquet call, the strict option should be set to False whenever the validation option allows that.
That would allow reading data that was serialized with a previous version (which then might require validation).