concurrency agent not working reliably with multiple clusters

With multiple clusters Openwhisk can create multiple containers for a concurrent action even with minimal traffic. Invocations in a cluster will create a new container in the same cluster. If invocations (actual invocation to debug vs. wskdebug polling) happen in different clusters, there will be 2 containers and they won't see each other, and wskdebug will fail to catch any invocations. This has been observed as a new behavior.

[Some internal discussion](https://cq-dev.slack.com/archives/C68UJPDM5/p1590554686290700).

### Workaround

Use another agent, either:
1. `--ngrok`
2. `--disable-concurrency` (available since 1.3.0, uses the activation db polling agent)