Add a sparseTreeLevel, make it the default sparse level

[willow-ahrens]

We can use SortedDict as an initial implementation. In particular, we should use https://juliacollections.github.io/DataStructures.jl/latest/sorted_containers/#Iteration-and-Token-Manipulation here.

We should also add some nonsequential indexing tests, as well as more principled nonsequential access tests for any levels that support it.

[willow-ahrens]

This level is not a critical piece of infrastructure for benchmarks in papers, but it is a high-priority feature for practical usability, so we are deferring this to a 1.0 milestone