Looper

Changed
- Use independent `peppy` package, replacing `models` module for core data types.
- Integrate `ProtocolInterface` functionality into `PipelineInterface`.

Changed
- Correctly count successful command submissions when not using `--dry-run`.

Changed
- No longer falsely display that there's a submission failure.
- Allow non-string values to be unquoted in the `pipeline_args` section.

Added
- Add `--lump` and `--lumpn` options
- Catch submission errors from cluster resource managers
- Implied columns can now be derived
- Now protocols can be specified on the command-line `--include-protocols`
- Add rudimentary figure summaries
- Simplifies command-line help display
- Allow wildcard protocol_mapping for catch-all pipeline assignment
- Improve user messages
- New sample_subtypes section in pipeline_interface

Changed
- Sample child classes are now defined explicitly in the pipeline interface. Previously, they were guessed based on presence of a class extending Sample in a pipeline script.
- Changed 'library' key sample attribute to 'protocol'

Added
- Add support for implied_column section of the project config file
- Add support for Python 3
- Merges pipeline interface and protocol mappings. This means we now allow direct pointers to `pipeline_interface.yaml` files, increasing flexibility, so this relaxes the specified folder structure that was previously used for `pipelines_dir` (with `config` subfolder).
- Allow URLs as paths to sample sheets.
- Allow tsv format for sample sheets.
- Checks that the path to a pipeline actually exists before writing the submission script.

Changed
- Changed LOOPERENV environment variable to PEPENV, generalizing it to generic models
- Changed name of `pipelines_dir` to `pipeline_interfaces` (but maintained backwards compatibility for now).
- Changed name of `run` column to `toggle`, since `run` can also refer to a sequencing run.
- Relaxes many constraints (like resources sections, pipelines_dir columns), making project configuration files useful outside looper. This moves us closer to dividing models from looper, and improves flexibility.
- Various small bug fixes and dev improvements.
- Require `setuptools` for installation, and `pandas 0.20.2`. If `numexpr` is installed, version `2.6.2` is required.
- Allows tilde in `pipeline_interfaces`

Added
- Add new looper version tracking, with `--version` and `-V` options and printing version at runtime
- Add support for asterisks in file paths
- Add support for multiple pipeline directories in priority order
- Revamp of messages make more intuitive output
- Colorize output
- Complete rehaul of logging and test infrastructure, using logging and pytest packages

Changed
- Removes pipelines_dir requirement for models, making it useful outside looper
- Small bug fixes related to `all_input_files` and `required_input_files` attributes
- More robust installation and more explicit requirement of Python 2.7