[imaging/database] Clean-up `mri_protocol` tables

[maximemulder]

After discussing imaging with @nicolasbrossard and @ridz1208 this morning, we agreed the mri_protocol tables could use some clean-up :

• In the mri_protocol_group_target table, the fields ProjectID, CohortID and Visit_label should have a UNIQUE constraint.
• In the mri_protocol table, the field Scan_type should have a foreign key to the mri_scan_type.ID field (the current default to 0 is also quite suspicious).

Eventually, we could rename some fields to respect more accurately LORIS naming conventions.