Add search support for datasetName and datasetID

[MortenHofft]

Prompted by gbif/portal-feedback#3006 and gbif/portal-feedback#3026

The desire in above issues it to have a way to search for occurrences within or across datasets using datasetName https://dwc.tdwg.org/terms/#dwc:datasetName and datasetID https://dwc.tdwg.org/terms/#dwc:datasetID.

And to allow multiple values when the publisher delimits them by |.

So if a publisher provides an occurrence withdatasetName: projectA|projectB then it should be possible in the API to find that occurrence by adding a filter like datasetName=projectA (a keyword filter in ES terminology I guess?)

I do not believe we need a suggest endpoint for this.

[rondlg]

Marvelous. Thankyou.

[MortenHofft]

I just realise that I haven't address these questions:

Current data that use pipes in a different way
The documentation currently do not say anything about using a pipe separator to split multiple values for datasetName.
So we have data that use |, but where it probably isn't intended as 2 names as such but as a hierarchy. E.g. Natural History Museum Denmark | Herpetology Collection
https://api.gbif.org/v1/occurrence/search?occurrenceID=5d867879-6e7d-4463-8773-e005a6a9731d

Would it make sense (and feasible) to evaluate how much data would be misinterpreted this way and try to get it changed at source?

Current response is a string
This issue suggest that we interpret this an array of keywords for search. Should the response format stay the same as now: a string. Or should it be changed to an array of strings? And where do we draw the line for API stability?

[rondlg]

If it's not a heavy lift to check for you to check for the delimiter then I'd say go for it. That said, certainly for us using a pipe to delimit each value would be the preferred way to handle it. We use that for other dwc fields and so isn't a terrible ask for us to fix at source (if we haven't already)!

Oh, and yes an array of strings please if possible.

[MortenHofft]

It would be considered a breaking change if we changed the response format from datasetName: "projectA | projectB" to datasetName: ["projectA", "projectB"]. As in projects using the code would be likely to fail.

So if we were to return an array in the response, then we would need a new field. Instead we will go for keeping the response format as a string. But it will be indexed as an array and hence be searchable as such. Meaning that you will be able to search for datasetName: projectA and get the occurrence back. Consumers that need the atomised values for display will need to split on | themselves.

[rukayaj]

Anything I can do to help with this?