Toil

- Don't error if job hasn't been submitted yet (resolves 1666)
- Match doc to reality for --preserve-environment
- CWL: symlink jobStore files instead of copying
- When killing tasks, make sure to reap child processes in the process groups
- Better error checking in piped docker commands
- Fix case where a job has been put in currentjobs, but not yet dispatc…
- Use hard links instead of copied files on import to file job store
- CWL: Add support for InitalWorkDirRequirement
- Set proper value for service job preemptable requirement
- Use chunked transfer in AWS job store importing
- Lower timeout for domain binding in Toil Clean
- Add `import os`.

* Updated documentation at http://toil.readthedocs.io/en/latest/
* Improved Azure support
* Remove upper limit on size of jobs in AWS job store
* CWL workflows can now take advantage of Toil Autoscaling
* CWLToil bug fixes and enhancements
* Toil Appliance updated to Mesos version 1.0.1
* Improve scalability of Mesos batch system
* Better error message for unfulfilled promises
* Fix handling of checkpointed jobs
* Fix service job deadlock detection
* Allow disabling of hot deployment mechanism

- New Toil landing page at http://bd2kgenomics.github.io/toil/
- Tags all Toil-provisioned instances with an 'Owner' tag to improve cost tracking
- Allows users to arbitrarily tag their Toil-provisioned instances
- Adds a new Torque batch system class
- Improved documentation
- Support for piping commands within DockerCall
- Improved cluster usage statistics
- Various bug fixes improving support for large file exporting, the Slurm batch system, and general Toil robustness

Hotfix release for 3.5.x
- Fixes job state discovery race
- Fixes race between job store file exporting and caching
- Fixes bugs in Slurm batch system

Hotfix release for 3.5.x
- Fixes import error raised when 'AWS' extra not installed

- Toil now includes a native EC2 provisioner for launching and controlling Toil clusters.
- Toil now has the ability to autoscale - if enabled, Toil will increase or decrease the number of worker nodes in its cluster depending on the current demands of the workflow
- Log output now refers to each Job by name.