This switches planemo in a subtle way from to depending more on the sturctured data json than the XUnit - this will be benefical for non-Galaxy tool testing when XUnit will be a product of the reporting stuff and not an input to it. This refactoring/reworking should allow most of planemo/galaxy/test/actions to be used with the newer planemo-driven testing stuff.

Adds tests and doc strings.

``Runnable``s are things such as  Galaxy tools, Galaxy workflows, CWL tools, CWL workflows and ``Engine``s provide operations over ``Runnable``s. There are two Engines right now - cwltool and Galaxy and one command that uses the Engine abstraction (``run`` - formerly ``cwl_run``).

Exit code handling is improved to so fixed exit codes result when unknown runnables are encountered or when a runnable is matched with an incompatible engine. A dependency on ``aenum`` is added to get enable nice enum functionality in planemo to describe runnable types.

The ``Engine`` abstraction implements running Galaxy and CWL tools so far (with either Galaxy or cwltool) but this commit contain a lot of infrastructure for running Galaxy tools and workflows of either type. In particular, workflows will now be uploaded as needed and IDs tracked, they can be executed, and what remains is to collect outputs. The generalized run command implements galaxyproject#449 and galaxyproject#364 - at least in some initial fashion.

Initial infrastructure for a generic test command across engine and runnable types is also present (xref galaxyproject#448). If ``planemo test`` encounters a runnable type other than galaxy_tool and directory - such as a cwl tool, a workflow, etc... it will use a new planemo engine-driven testing procedure. To do this an engine "job" is executed, the results collected, and the results "checked". Currently, Galaxy workflows do not have their results collected and the only "check" available is exact non-File matches and sha1 checksums as used by ``cwltest.py``. A complete suite of checks matching what is available in Galaxy will hopefully be implemented shortly.

This new engine-driven testing produces a structured test data structure that should match what is generated by Galaxy regardless of which engine is used. The upshot of this is that existing test report infrastructure in planemo should be reusable. ``planemo test_report`` and ``planemo share_test`` should work for any ``Engine`` and any ``RunnableType``.

Despite not yet using ``Engine``s infrastructure, the ``serve`` command and related infrastructure for Galaxy has been updated to consume runnables instead of paths. An upshot of this is that serve has been updated to allow serving workflows (either ``.ga`` files directly or Format 2 Workflows through the new gxformat2 library).

In order to serve, run, and test gxformat2 workflows - planemo now depends on the gxformat2 library (https://github.com/jmchilton/gxformat2). The depenedency on bioblend has been constrained to be newer so that workflow invocation client stuff can be used.

The commit also contains general improvements for CWL support - in particular the cwl-runner implementation driven by planemo is much enhanced and now handles (in at least a minimal fashion) all arguments required to be used by the conformance test code. This commit also introduces a test script ``scripts/test_confromance_cwl.bash`` that runs the common workflow language conformance tests against Planemo/Galaxy. Fixes galaxyproject#453.

This commit also includes a variety of improvements to ``run`` and ``serve`` tests.

 - Refactor serve test cases to allow more reuse.
 - Fix profile test to allow rerunning on the same machine more than once.
 - Docstrings.