Adaptive concurrent downloads

Implement an adaptive concurrent download strategy to account for varying internet speeds.

- The current threaded downloading works fine on the cloud server (defaults to 50 threads downloading concurrently, which seems to be about right), but running locally in an area with slow internet often results in timed-out tiles.
- Ping suggested http://docs.python-requests.org/en/master/

No file, test, or entry point is named; start by locating the current threaded tile downloader and review the linked Python Requests documentation. Define the adaptive behavior for slow and fast connections, then verify that downloads avoid the reported timeouts while retaining suitable concurrency on the cloud server.

Written by the indexing model from the issue text.