Avoid deadlock if worker crashes with an exception

New major release

Add hvd.local_size() API

Add Keras ImageNet training example and LearningRateScheduleCallback

Add -stdlib=libc++ to CFLAGS for MacOS

Modify LRWarmupCallback to scale LR from `initial_lr / hvd.size()` to…

… `initial_lr`

Bugfix: Initialize MetricAverageCallback variables

Add support for boolean broadcast() and allgather()

First-class Keras support

Add support for new tf.sysconfig.get_compile_flags() and tf.sysconfig…

….get_link_flags()