Add DocumentHeaderModel from crm.document_headers

https://eaflood.atlassian.net/browse/WATER-4292

> This is part of a series of changes to add the ability to identify the licence holder for a licence

Arghhhhh! 😱🤬😩

We've not long ago added 3 new models to this project

- [LicenceDocumentModel](#618)
- [LicenceRoleModel](#629)
- [LicenceDocumentRoleModel](#630)

We then [linked the LicenceModel to LicenceDocumentModel](#632). All this was based on a first pass of what we thought the [water-abstraction-ui](https://github.com/DEFRA/water-abstraction-ui), [water-abstraction-service](https://github.com/DEFRA/water-abstraction-service) and [water-abstraction-tactical-crm](https://github.com/DEFRA/water-abstraction-tactical-crm) apps were doing to determine the licence holder. It reflected what we found in the tables behind the 3 models we added as well.

We'd just [coded up a service](#639) to set the return requirement setup session with the name of the licence holder and begun testing when we found an issue in testing.

We came across a licence where neither the company or contact reflected what we saw in the UI for the licence holder! In the UI beneath the licence holder name is a 'View licence contact details' link. We were able to figure out that the ID it uses is for a `crm.document_headers` record. And in there we found (surprise-surprise!) another JSONB field holding a bunch of data, including the licence holder name we were seeing on screen.

Another dive into the legacy code revealed we'd taken a wrong turning. When the page is requested there is logic that grabs records from the tables we created models for. But they are not a used. In fact, not even the `crm.document_headers` record is used, other than to provide an ID for the link.

What seems to happen (we could be wrong again!) is that the `permit.licence` table is queried and its `licence_data_value` field is extracted. If you haven't guessed already, this is another JSON field. Only this one seems to be a complete 'dump' of everything imported from NALD for the licence. This data gets passed to `src/lib/licence-transformer/nald-transformer.js` which transforms it from the NALD structure into a 'licence'. This includes a `licenceHolderFullName` property which is eventually what appears in the UI.

We have absolutely _zero_ intent to start importing and transforming raw NALD data just to get the licence holder name. Thankfully, in the follow up testing we've been doing the information held in `crm.document_headers` has always matched what we see in the UI.

So, this is the start of our second attempt to crack licence holder name starting with adding the `DocumentHeader` model.