Reintroduce sync client

[tomchristie]

Builds on #713 and reintroduces a sync Client.

>>> import httpx
>>> httpx.get('https://www.example.com')
<Response [200 OK]>

🎉✨ TA-DA! ✨🎉

We're using urllib3 here as the backing service for our sync calls.

Potentially we could later switch away to using our own unasync'ed dispatcher. (Although it's actually feasible that we might just want to stick with urllib3 for the sync case.)

If we wanted to tackle this as a set of incremental PRs we would want:

• Switch Dispatcher to AsyncDispatcher throughout.
• Introduce a SyncDispatcher class.
• Introduce a Proxy config class.
• Split the client into the two sync and async cases, with a base class of BaseClient.
• Update StreamContextManager to handle both sync and async cases.
• Switch the top-level API back to sync.

Also still TODO here:

• Tie in timeout and pool limits config properly.
• Introduce a WSGIDispatcher class.
• Introduce a URLLib3ProxyDispatcher class.
• Plenty of docs updating.
• Plenty of test updating.
• Wrap urllib3 exceptions (Need to cover pool.urlopen and conn.stream methods)

[tomchristie]

Now closing in favour of #735