Spatial Indexes may be slow for certain queries

Depending on the table and the query, performance might not be great for Spatial Indexes.
Problematic situations:
- Large selections
- Filters that select large portions of rows are slower through Spatial index
- Large geometries
- Tables with large polygons (especially ones that cross axis boundaries) tend to be slower

There are a couple optimization options, some are harder than others, and some may break compatibility between dolt versions.
Potential Optimizations:
- Level Pruning
- For every query, we generate 65 ranges; many tables don't have objects that large,
- Covering Indexes
- Spatial Indexes are non-covering, which forces an extra level of indirection, slowing down lookups
- Geometry Decomposition
- Can further minimize the number of false positives by breaking down large objects into smaller ones

The issue names no files, tests, or entry points. Start by locating the spatial-index implementation and measuring the listed large-selection and large-geometry cases; done requires selecting and validating an optimization without breaking compatibility between Dolt versions.

Written by the indexing model from the issue text.