Don't report a cluster as ready to work until node connection protocol has completed

[jtfmumm]

Previous to this commit, we were reporting ready to work at
a boundary as soon as the TCP connection was established.
Since we are not actually ready to work until all boundary
connect protocol messages have been received, this changes that.

Closes #1763

[slfritchie]

There are two tests failing, test_restart_pony and test_restart_machida in testing/correctness/tests/restart_without_resilience/restart_without_resilience.py. I can duplicate the failures by running make integration-tests-testing-correctness-tests-restart_without_resilience verbose=true debug=true on my MacBook.

The last 5 lines of the initializer's output in each test are:

Muting TCPSource
Muting TCPSource
Muting DataChannel
The same Initializable reported being ready to work twice
This should never happen: failure in /Users/scott/s/src/wallaroo/lib/wallaroo/core/initialization/local_topology.pony at line 1637

The full stdout from the initializer process for both tests is at https://gist.github.com/slfritchie/a3dc74b579708ba2a68b50403b4145d3

[jtfmumm]

Ok I think I understand that error. I'll fix that tomorrow morning. Thanks.

[jtfmumm]

@slfritchie That error was because I was overreaching in my initial attempt. I've scaled it back to just what we need and tests are passing now.

[jtfmumm]

@SeanTAllen This makes it so that we now wait for the boundary connect protocol process to complete before we say we're ready to work.

[SeanTAllen]

@jtfmumm i updated the title

[jtfmumm]

@SeanTAllen Ok

[slfritchie]

The fix works well: it doesn't have any transient failures that I can see in a few hundred attempts. @SeanTAllen any additional review?

[SeanTAllen]

@slfritchie nope. i merged.