Better warnings for incorrect batch size in replay buffer

[djl11]

Here is a minimal example:

from tf_agents.replay_buffers import tf_uniform_replay_buffer
import tensorflow as tf

batch_size = 8
counter = 0

data_spec = {'observations': tf.TensorSpec([3], tf.float32, 'obs'),
             'actions': tf.TensorSpec([3], tf.float32, 'actions')}

replay_buffer = tf_uniform_replay_buffer.TFUniformReplayBuffer(
    data_spec=data_spec, batch_size=1, max_length=1000)

while True:
    print(counter)
    counter += 1
    replay_buffer.add_batch({'observations': tf.zeros((batch_size, 3)), 'actions': tf.zeros((batch_size, 3))})

This sometimes runs properly, and other times leads to:

Process finished with exit code 139 (interrupted by signal 11: SIGSEGV)

This is hard to debug, and the batch size mismatch could be hard to spot in larger project files. I think an error should be thrown if you try add add an incorrectly sized batch :)

[leroidauphin]

This looks like a really interesting problem, so I attempted to reproduce it to see if it could be fixed. I can reproduce this in tf-agents 0.5.0, exactly as you describe, but I cannot reproduce this in the nightly build. Are you still seeing this SIGSEGV when using the nightly build?