Normalize dataset field types

[wslulciuc]

When dataset metadata is collected, the backend API persist the type for a dataset field using the raw string provided by the caller. But, a field type varchar is the same as VARCHAR. To account for the case insensitivity, the backend invokes a toUpperCase() on the type to avoid creating a new dataset version if the field types are essentially equivalent.

So, the goal of this issue is to begin the discussion on what normalization should be applied on the backend for dataset field types. I think it's reasonable to normalize on all uppercase, therefore providing a consistent format on how types are persisted.