Make sure RE is closed

[dominikl]

The info command creates a RenderObject which loads a RenderingEngine but doesn't close it. When you run it on a large Plate you soon hit an OutOfMemory exception. This PR closes the RenderingEngine again.

The PR also adds a thumb command which generates rendering settings (if there aren't any yet) and creates the thumbnails. Without it one has to run first render info and then test --thumb.

[dominikl]

Note: It makes the info command much slower, 2-3 times, unfortunately. But leaving the rendering engines open isn't really an alternative.

[dominikl]

I'm running this currently in pilot-idr0093. It takes ages, and needs to be followed by another run of render test --thumb to create the thumbnails. This will take a long time too, because again it needs opening and closing rendering engines. Hence, I'll add another command 'thumb' which combines these two steps, so that a rendernig engine only has to be created once per image.

[jburel]

Note: It makes the info command much slower, 2-3 times, unfortunately. But leaving the rendering engines open isn't really an alternative.

I don't think starting a rendering engine to get the rendering setting info is the way to go when the settings exist.
it is an overkill

[dominikl]

Could improve the RenderObject in another PR, add an option to load the settings without opening a rendering engine. But at least for now the PR makes sure the open rendering engine is closed again. Having the option to start a rendering engine is necessary too, in order to create rendering settings if there aren't any (I assume that's what it actually does).

[jburel]

When you don't have settings: yes, but when they exist it is not needed.
It is similar in a way to https://github.com/ome/omero-py/blob/master/src/omero/gateway/__init__.py#L8644

[jburel]

info and thumb commands now close the RE.
Since this PR is about closing RE, could you also check the tests?, e.g. assert_target_rdef

[dominikl]

Sure. I'll also try to fix the RenderObject in this PR, so that it doesn't unnecessarily open a rendering engine.

[dominikl]

I removed the last commit again. A separate thumb command as wells as @jburel 's suggestions should be part of a separate PR. So this PR is really only about closing the rendering engine, so that the PR can be merged and the leak fixed.

[joshmoore]

LGTM. I can imagine that something like the "--keep-going" flag that we have elsewhere (import.py) could be used to determine a throw or a ctx.err.