LRU cache now stores per jumpinfo per field, instead of per file.  Add
ability to forward scan requests to another server Optimise extraction
tasks and serialisation of requests when many ranges are requested.