Add resuming capabilities to tactile pushing example

Would be useful to test (if not already) a scenario where the training is cut off and resumed from a checkpoint, and a scenario where inference is performed by loading a model that is trained and saved from before.

_Originally posted by @mhmukadam in https://github.com/facebookresearch/theseus/pull/55#pullrequestreview-869120170_

Start by locating the tactile pushing example and its training and inference entry points. Check how checkpoints are currently saved or loaded, then add coverage for resuming interrupted training and for inference from a previously saved model; done means both scenarios run successfully.

Written by the indexing model from the issue text.