Add client-side caching of GraphQL requests.

This probably already exists in the most basic format. Look into improving it, especially for subqueries e.g After fetching all the data for a particular indicator, if a user then wants to see a subset of that same data relating to a specific district, don't run the query again. Instead, fetch from the cached results.

Locate the client-side GraphQL request and caching implementation, then trace the flow for fetching an indicator and a district subset. Verify whether the subset request repeats the network query; done means matching subset data is served from cached results without another request.

Written by the indexing model from the issue text.