Support rendering without randomizing camera position, body or clothing

[rib]

glimpse-cli.py is primarily for generating training data where we want to see lots of variation to the camera viewing angle, camera height, distance from the body, body and clothing.

Sometimes though it would be useful to render out one or more motion capture sequences so we could convert a sequence into a glimpse viewer recording that we can load as synthetic test data. In this case if the camera were to move at all then it needs to have temporal consistency (e.g as if someone were holding a phone not perfectly still).

As a starting point it would be good to support rendering with a perfectly fixed camera, without randomizing the clothes and with a single body.

glimpse-cli.py and blender/addons/glimpse_data_generator/__init__.py probably need changing to allow the min/max viewing angle and height of the camera to be fixed. Currently the script will complain if they are equal because randrange(A, B) used internally will throw an exception if the given range is empty. Note: beware the way the viewing angle range is defined as two positive numbers for how far the camera may rotate left or right.

Maybe we could allow --clothing-step=-1 to be recognised as special so that we shouldn't randomize the clothes.

Maybe we could have a --body=N argument or support something like --body=Man0 for explicity choosing which body we want to render with.

This will involve modifying the validation of a few argparse arguments within glimpse-cli.py, then adding/updating a few corresponding Blender addon properties within blender/addons/glimpse_data_generator/__init__.py register() and then modifying the execute() method of the GeneratorOperator (also in blender/addons/glimpse_data_generator/__init__.py) to handle these properties to avoid randomization while rendering.