Index pruning

[silky]

It might be interesting to consider a feature to "prune" an index of certain items; to reduce the RAM requirements for hosting the dataset.

Perhaps it's useful to have the "full" index for a certain stage of analysis, but at a later stage present only a relevant subset. It should be moderately straightforward to just delete any un-needed items from the index and wildly cut down on the resulting file size.

How to know it works

Simple test case

• Have a minimal geojson dataset that we compute an index from.
• It should consist of 3 items
• Use it to compute indicies of size 1, 2, and 3.
• Check that we can trim the 3 index to equal the 2 index,
  • and the 2 index to be the 1 index,
  • and that we can trim the 3 index to be the 1 index.
• That should be enough.