Get elevation data from a pure Rust crate that only handles some

[dabreegster]

GeoTIFFs. #82

This PR switches from Docker + Python/GDAL-based https://github.com/eldang/elevation_lookups to pure Rust https://github.com/dabreegster/elevation. The benefits are:

• Portability. In the span of a year or two, the Docker + GDAL based environment broke on newer Ubuntu, making it a chore just to re-run. Pure Rust means everything compiles anywhere with no external dependencies.
• Simplicity. The code is just bilinear interpolation on top of https://github.com/pka/georaster/. The simplifying assumption that removes the need for GDAL is that the input is a GeoTIFF in EPSG:4326. gdalwarp -t_srs EPSG:4326 kc_2016_lidar.tif king_county_2016_lidar.tif as a one-time step is all it takes to transform the Seattle file, and the UK one was already in the right coordinate system.
• Speed. The South Seattle elevation step drops from 4.2s to 1.67s.

A future step could look for SRTM / NASADEM geotiff files and do any coordinate transformation necessary for use anywhere. And separately, use a pure Rust approach for interpolating from GeoJSON contour files.

Validation was regenerating Seattle and UK maps and checking the elevation / steep street layers in the simulation mode. The results are not exactly the same, but close enough. Regenerating all maps again now, will merge when that's done.

[dabreegster]

Edit: between this and yesterday's switch to osm.pbf for imports, I can regenerate all 156 cities on my fastest laptop in about 15 minutes! I don't quite remember what it was before, but I'm quite sure this is a nice improvement