fix: re-establish watcher dropped connections

From time to time, the watcher connection will drop, this will leave the
workflow hanging forever to catch an event that will never arrive.
Upon exception catching of:

```
Unexpected error: ("Connection broken: InvalidChunkLength(got length b'', 0 bytes read)", InvalidChunkLength(got length b'', 0 bytes read))
```

We re-establish the watcher's connection.
We also configured a timeout of 1m for each watcher.

Once, the informers are implemented in the kubernetes python library it
will be more robust to switch that. This is tracked here:
kubernetes-client/python#868.

Signed-off-by: Sébastien Han <[email protected]>