Add overall timeout support

[karpetrosyan]

Todo

• Update documentation

Hi! It seems we have received many requests for adding total timeout support.

In each place where we are using the operation (read, write, pool, connect) timeout, we should take the minimum value of the operation timeout (timeouts['pool'], for example) and the total timeout value (timeouts['total']). We should also measure the time of each operation and adjust the value of the total timeout accordingly, so we always have the correct total timeout value within the scope of the request.

I'v used the context manager syntax to simplify time measuring, also have added a simple api level test for total timeout, as we don't have any tests for the concrete timeouts

[karpetrosyan]

@tomchristie Any thoughts?

[tomchristie]

Thanks @karpetrosyan...

Any thoughts?

• This is a fantastic piece of functionality to add yes, much needed.
• Could we get a docs update.
• Related work that's become apparent to me is that read/write timeouts are the wrong API... a single socket timeout would be a better fit.

[karpetrosyan]

I guess that's all from our side. The remaining documentation will be available in the HTTPX docs.