Pydra

- reorganization of the `State` class, fixing small issues with the class
- fixing some paths issues on windows os
- adding osx and window sto the travis runs (right now allowing for failures for windows)
- adding `PydraStateError` for exception in the `State` class
- small fixes to the hashing functions, adding more tests
- adding `hash_dir` to calculate hash for `Directory` type

- passing `wf.cache_locations` to the task
- using `rerun` from submitter to all task
- adding `test_rerun` and `propagate_rerun` for workflows
- fixing task with a full combiner
- adding `cont_dim` to specify dimensionality of the input variables (how much the input is nested)

* adding sphinx documentation
* moving from `dataclasses` to `attrs`
* adding `container` flag to the `ShellCommandTask`
* fixing `cmdline`, `command_args` and `container_args` for tasks with states
* adding `CONTRIBUTING.md`
* fixing hash calculations for inputs with a list of files
* using `attr.NOTHING` for input that is not set

supporting tuple as a single element of an input

- fixing: nodes with states and input fields (from splitter) that are empty were failing

* big changes in `ShellTask`, `DockerTask` and `SingularityTask`
- customized input specification and output specification for `Task`s
- adding singularity checks to Travis CI
- binding all input files to the container

* changes in `Workflow`
- passing all outputs to the next node: `lzout.all_`
- fixing inner splitter

* allowing for `splitter` and `combiner` updates

* adding `etelementry` support