Switch WMS router to use async

[mpiannucci]

The approach for this is to delay dask chunk downloading for GetMap requests to a single dask compute call that we call in asyncio to_thread(). The rest of the methods simply use to_thread() for now but there is a path to support a more featureful implementation.

This also introduces clipping for GetMap requests to decrease the cpu/memory pressure on the rendering process, as well as decreasing the amount of chunks that need to be downloaded.

[mpiannucci]

Notes

• Dont have two thread pools (dask and asyncio), or change their sizes.