Reregister existing framework and support escape of slash in task names

[kohend]

I found out that when I tried to reregister an existing framework I received a mismatch of the framework ID in different fields:
Failed to validate scheduler::Call: 'framework_id' differs from 'subscribe.framework_info.id'
Since the subscribe call sets the framework ID to null.
This sets it to the framework ID given if defined, and prints out the error output from Mesos upon failure to register.
In addition, since we wanted to use task names with slashes to match the marathon groups we're using, I added an escape of the task name to underscores, like marathon does internally.

[tobilg]

Just to understand this correctly, with that change you ideally wan to handle re-registrations "only", while the scheduler is still running? self.frameworkId is never set from outside, so if the scheduler gets restarted, I guess there's no way to recover (yet)

[tobilg]

Thanks for the PR! As said in my review comment, I'd like to understand the use case better. As the frameworks don't persist their data in ZK yet, I think there's currently no way to restart a scheduler... This is yet to be fixed.

[kohend]

I wrote code to persist the framework ID in ZK, I put it in the "boilerplate" end of things (currently in our internal git, waiting for code review), I understand that task information needs to be persistent as well to allow a takeover of a restarted scheduler, but this is a step in the right direction.
It was impossible to develop when each restart registered a new framework.

[tobilg]

Merged, thanks! Let's have a chat about the ZK stuff if you like. I'm both in the Mesos and DC/OS Slack channels.

[tobilg]

I released the version 0.2.9 just now.