Gridded data

Nobody has claimed this yet.

Tippecanoe's gridded data support is limited to `--grid-low-zooms`, which requires a lot of guesswork to choose an appropriate detail and maxzoom, uses the tile grid rather than the natural grid of the data, and drops rows and columns instead of interpolating values.

There should be a `tippecanoe-grid` command that is roughly equivalent to `gdal_polygonize` (and perhaps also `gdal_contour`) but instead of outputting a single set of polygons for the grid, outputs a series of grids appropriate to different zoom levels, with `"tippecanoe": { "minzoom": …, "maxzoom": … }` set so they show up at the appropriate zoom levels.

For manageable tile sizes, the grid should be approximately 128x128 in each tile, so if the grid unit is, say, 1000 feet, it should choose zoom level 10 for the full-detail zoom level, because `360 / (2^10) / .00000274 / 128` is about 1000. Each lower zoom level should then cut the number of cells in each dimension in half.

Major uncertainties:

* What file format should it accept? Is there a format that is each for people to get their data into, sufficiently general, and not too much work to parse?
* How much projection support is needed? Ideally it would handle the natural projection of the gridded data instead of requiring that it already be reprojected to WGS84 or Web Mercator.
* Does it need to be able to reconstruct the grid of data that has already been polygonized?

Start by reviewing the existing --grid-low-zooms behavior and the proposed tippecanoe-grid command. Resolve the open questions about input format, projection support, and reconstructing polygonized grids before implementation. Done means producing zoom-appropriate grids with tippecanoe minzoom and maxzoom metadata and approximately 128x128 cells per tile.

Written by the indexing model from the issue text.