Don't allow listing indexes via the `retrieveDocument` API

[tripott]

We would need to:

• Prevent new documents from having the _id _index, either by rejecting those documents, or mapping the string _index to another internally managed string. Affected apis: createDocument bulkDocuments
• when a retrieveDocument request arrives, either reject the lookup for _index or map the string similar to aforementioned mapping
• IF the mapping approach is taken, the document's _id would need to be reverse mapped back to the string _index on any retrievals. Affected apis: retrieveDocument, listDocuments, queryDocuments
• add a test case for each

The easier of the two is just to reject any documents with an _id set to _index, but is arguably leaking impl details to the consumer.

Background

Related Issue: hyper63/hyper#486

So something I discovered today that probably should not be allowed?

If I have a hyper data service, I can retrieve all of the indexes that exist on the database by calling GET /data/default/_index. This gets handled by the "get document by id" route, which calls retrieveDocument on the adapter.

Ultimately the CouchDB adapter does:

asyncFetch(`${config.origin}/${db}/${id}`,...)

with _indexas the id.

Of course, there isn't a document with _id _index, BUT that does map to the route on the CouchDB API that lists all of the indexes for the DB: https://docs.couchdb.org/en/stable/api/database/find.html#get--db-_index. And since the data port says anything can be returned from the retrieveDocument API, hyper simply returns that response and you get something like:

{
	"total_rows": 1,
	"indexes": [
		{
			"ddoc": null,
			"name": "_all_docs",
			"type": "special",
			"def": {
				"fields": [
					{
						"_id": "asc"
					}
				]
			}
		}
	]
}