Introduce EndpointResult

[vkostyukov]

Currently, Endpoints are backed by Option[(Input, Rerrunable[Output[A]])]s. There is nothing wrong with that type and it provides a great insight on what happens in the endpoint, but there is a way to make it 1) simpler and 2) faster.

Let's flattenOption[Tuple2[_, _]] into a new type EndpointResult (ADT with two cases: matched vs. skipped) and save some allocations on option creations.

In addition to that this PR cleans up a couple of places that I think were poorly structured:

1. All the testing methods on endpoint's result are now explicit (no need for implicit syntax class over the Option).
2. All blocking methods are renamed (deprecated) to awaitX variants to visually signal the reader that the call is blocking. They also take optional Duration, which comes in handy in tests.

UPDATE: I decided to give up on a mutable EndpointResult for now since it doesn't buy much of performance improvement but tradeoffs immutability. The most recent benchmark run is here (https://gist.github.com/vkostyukov/e54092f843280dc3f8930c1d359db342). TL;DR EndpointResult[_] performs better or same as Option[Tuple2[_, _]].

[codecov-io]

Current coverage is 76.85% (diff: 74.04%)

Merging #707 into master will decrease coverage by 0.36%

@@             master       #707   diff @@
==========================================
  Files            33         34     +1   
  Lines           641        661    +20   
  Methods         618        635    +17   
  Messages          0          0          
  Branches         23         26     +3   
==========================================
+ Hits            495        508    +13   
- Misses          146        153     +7   
  Partials          0          0          

Powered by Codecov. Last update 136dbe4...f68f0ae