Add support for chaining State Partition -> Stateless Partition

[jtfmumm]

This PR builds on Nisan's PR #1662. It solves the problem he identified for this topology layout.

Closes #1663
Closes #1110

[nisanharamati]

Status as of commit 7a4962f:

• the first issue (error thrown from router.pony:348) appears fixed
• the second issue (error thrown from typed_route.pony:86) is still occurring
• there is a new issue: incorrect output intermittently appears in test runs where the number of workers is >=3 (I've seen it in runs with 3, 4, 5, 9, and 12 workers) for state_partition_parallel_stateless_state_partition_app (run make clean && make test debug=true then repeat make test until you encounter the failure with the error below)

  expected final sum to be:
  [(33, 198), (34, 200), (33, 196)]
  but received:
  [(1, 2), (2, 20), (1, 16), (2, 16), (3, 22), (3, 20), (1, 12), (2, 30), (3, 30), (4, 30), (5, 30), (4, 20), (5, 26), (6, 38), (7, 56), (4, 22), (5, 28), (8, 56), (9, 56), (10, 68), (6, 34), (7, 52), (8, 70), (9, 70), (10, 70), (11, 74), (6, 48), (7, 66), (8, 66), (9, 66), (10, 66), (11, 66), (12, 74), (13, 74), (14, 80), (15, 92), (16, 110), (11, 70), (12, 70), (17, 110), (18, 110), (13, 88), (14, 106), (15, 106), (16, 106), (17, 112), (19, 128), (12, 84), (13, 102), (14, 102), (15, 102), (16, 102), (17, 102), (18, 108), (19, 114), (20, 132), (20, 128), (21, 128), (22, 134), (23, 134), (24, 146), (18, 112), (19, 112), (20, 124), (21, 142), (22, 142), (23, 148), (25, 146), (21, 132), (22, 138), (23, 156), (24, 156), (25, 156), (26, 156), (26, 152), (27, 164), (28, 182), (24, 148), (25, 148), (26, 154), (27, 160), (28, 178), (29, 196), (30, 196), (31, 196), (29, 182), (30, 182), (31, 194), (27, 174), (28, 192), (29, 192), (30, 192), (31, 198), (32, 198), (33, 198), (32, 194), (33, 194), (34, 200), (32, 196), (33, 196)]

Note1: the error prints out the entire received set, while the validator is testing the final value in each partition.
Note2: there are repetitions of values in the output... Since the logic is max -> double -> max, this shouldn't happen.

[nisanharamati]

Status as of c40aaf5:

• router.pony:348 bug looks fixed
• typed_route.pony:86 bug looks fixed
• topologies with any sort of state step following a Parallel Stateless step intermittently fail validation

[nisanharamati]

The intermittent validation failures are a result of the particular stateful computation: CountAndMax which keeps a record of the number of messages it has seen, and the highest message it has seen.
The reason this can intermittently fail validation in multi-worker runs with parallel stateless computations is that while the final value of CountAndMax is guaranteed assuming no data is lost in any partition, the intermediate values, and their order, isn't.

For example:
If the final value arrives _ before_ its preceding value: 1, 2, ..., 100, 99
Then the output sequence will be (1, 1), (2, 2), ..., (99, 100), (100, 100)
Whereas the expectation is: (1, 1), (2, 2), ..., (99, 99), (100, 100)

In order words, this test failure is a bug in the test design, and not in Wallaroo.

[nisanharamati]

This should pass all tests now.

[jtfmumm]

@SeanTAllen This is passing tests now. Since this now combines work from both @nisanharamati and me, I've assigned you as a reviewer.

[SeanTAllen]

Does this need to be squashed?

[jtfmumm]

@SeanTAllen My commits are logical commits with explanations. Not sure about all of @nisanharamati 's