How to use l1_distance with minibatch?

Hello, everyone and Developer of dynet.
Thanks for good library.

Now I trying to use `L1 loss function`.
I wrote my code which use L1 loss function with minibatch.
But I got following error.

```
terminate called after throwing an instance of 'std::runtime_error'
what(): Node || a - a ||_1 does not support batching but got fed batched tensor
```

What should I do in order to use L1 loss function with minibatch.
Is it possible to implement easily L1 loss function **directly** in my code?

I think it is difficult for me to implement L1 loss function to dynet library,
because I'm not familiar with inside of dynet library.

I'm glad if you help me or give some advice.
Thanks in advance.

Start by reproducing the reported runtime error when l1_distance receives a minibatch, then trace the L1 loss implementation and its batching support in DyNet. Done means determining whether minibatch use should be supported and documenting or implementing the agreed approach.

Written by the indexing model from the issue text.