Add keep-alive message between worker and scheduler

[mrocklin]

This is effectively a heartbeat, but much simpler and less frequent than
our current heartbeats

Fixes #2524

[lr4d]

Getting this error on the scheduler side:

2019-08-01 13:30:48,297 ERROR    <lambda>() got an unexpected keyword argument 'worker' (distributed.core)
Traceback (most recent call last):
  File "/mnt/mesos/sandbox/venv/lib/python3.6/site-packages/distributed/core.py", line 477, in handle_stream
    handler(**merge(extra, msg))
TypeError: <lambda>() got an unexpected keyword argument 'worker'

[mrocklin]

Thanks @lr4d . Handled

Also, what is a good time for the frequency here? Every minute? Every ten minutes? Every hour?

[lr4d]

I'd keep it below 5-10 minutes. For HAProxy the treshold for killing a connection on which no data is sent is 60 minutes, but I don't know what this may be for similar tools.

[lr4d]

This is appears to be working fine now for our HAProxy setup, I left the cluster alive for 6 hours and no disconnections or error messages took place.
Thanks @mrocklin