What is the feature and why do you need it:
All API calls done through this client today goes through urllib3, where the timeout is default to set to None, this means infinite timeout and is a well known pitfall with urllib3. Lost connection attempts can get stuck waiting forever, with no retries being done to reestablish the connection.

Quoting the requests-documentation which uses urllib3 under the hood:

Nearly all production code should use this parameter in nearly all requests. Failure to do so can cause your program to hang indefinitely:

There is a _request_timeout kwargs argument that can be set on each api-call, but that is both tedious, easy to forget and not type-checked due to use of kwargs-dictionarly.

Describe the solution you'd like to see:
Similar to how retries is configured, Configuration should accept a timeout-argument, this can be forwarded into urllib3s poolmanager where it uses this timeout for all operations where an explicit timeout is not provided.