Base search is unnecessary heavy

[tokee]

A basic search triggers

• A facet request with rows=0 (this is fine)
• A document request with rows=20&hl=true (fine) with a lot of fields (less fine as only a few are used)
• 20 separate document requests, one for each document shown, with extra fields for the documents (not fine, why not request those in the initial call and skip the 20 calls fully?)
• 20 separate image requests with time-proximity sorting (fine as this is needed), with a lot of fields (less fine as only warc+offset+width/heigh is needed)

Cleaning this up should improve responsibility and reduce search load.

[tokee]

After discussion with @thomasegense we agreed that

• It is a lot of requests and that number should be reduced
• Reducing the amount of fields is a minor issue, only to be done if it is easy
• Removing the 20 separate requests (#3 in the list above) should be done, either by collapsing it to a single request or by adding the field links_images to the initial call for documents