Add Lightning interface, Backbone classes, new classifiers, and data utils.

v0.1.6

Added

• PyTorch Lightning interface to MAML, ANIL, ProtoNet, MetaOptNet.
• Automatic batcher for Lightning: l2l.data.EpisodicBatcher.
• l2l.nn.PrototypicalClassifier and l2l.nn.SVMClassifier.
• Add l2l.vision.models.WRN28.
• Separate modules for CNN4Backbone, ResNet12Backbone, WRN28Backbones w/ pretrained weights.
• Add l2l.data.OnDeviceDataset and implement device parameter for benchmarks.
• (Beta) Add l2l.data.partition_task and l2l.data.InfiniteIterator.

Changed

• Renamed and clarify dropout parameters for ResNet12.

Fixed

• Improved support for 1D inputs in l2l.nn.KroneckerLinear. (@timweiland)