Python error reporting: Copy primary error to bottom of error output.

Fixes https://twitter.com/karpathy/status/765681445637533696

Example new output for a CPU-only binary that executes:

```
import tensorflow as tf
s = tf.Session()
with tf.device("/gpu:0"):
  s.run(tf.constant(1.0) + tf.constant(2.0))
```

is:

Traceback (most recent call last):
  File "baz.py", line 5, in <module>
    s.run([c])
  File "REDACTED/local/lib/python2.7/site-packages/tensorflow/python/client/session.py", line 717, in run
    run_metadata_ptr)
  File "REDACTED/local/lib/python2.7/site-packages/tensorflow/python/client/session.py", line 855, in _run
    raise RuntimeError('The Session graph is empty.  Add operations to the '
RuntimeError: The Session graph is empty.  Add operations to the graph before calling run().
(tf)vrv@vijay-linux:~/vrv.tensorflow/tf$ python baz.py
Traceback (most recent call last):
  File "baz.py", line 4, in <module>
    s.run(tf.constant(1.0) + tf.constant(2.0))
  File "REDACTED/local/lib/python2.7/site-packages/tensorflow/python/client/session.py", line 717, in run
    run_metadata_ptr)
  File "REDACTED/local/lib/python2.7/site-packages/tensorflow/python/client/session.py", line 915, in _run
    feed_dict_string, options, run_metadata)
  File "REDACTED/local/lib/python2.7/site-packages/tensorflow/python/client/session.py", line 965, in _do_run
    target_list, options, run_metadata)
  File "REDACTED/local/lib/python2.7/site-packages/tensorflow/python/client/session.py", line 985, in _do_call
    raise type(e)(node_def, op, message)
tensorflow.python.framework.errors.InvalidArgumentError: Cannot assign a device to node 'add': Could not satisfy explicit device specification '/device:GPU:0' because no devices matching that specification are registered in this process; available devices: /job:localhost/replica:0/task:0/cpu:0
	 [[Node: add = Add[T=DT_FLOAT, _device="/device:GPU:0"](Const, Const_1)]]

Caused by op u'add', defined at:
  File "baz.py", line 4, in <module>
    s.run(tf.constant(1.0) + tf.constant(2.0))
  File "REDACTED/local/lib/python2.7/site-packages/tensorflow/python/ops/math_ops.py", line 748, in binary_op_wrapper
    return func(x, y, name=name)
  File "REDACTED/local/lib/python2.7/site-packages/tensorflow/python/ops/gen_math_ops.py", line 71, in add
    result = _op_def_lib.apply_op("Add", x=x, y=y, name=name)
  File "REDACTED/local/lib/python2.7/site-packages/tensorflow/python/framework/op_def_library.py", line 749, in apply_op
    op_def=op_def)
  File "REDACTED/local/lib/python2.7/site-packages/tensorflow/python/framework/ops.py", line 2372, in create_op
    original_op=self._default_original_op, op_def=op_def)
  File "REDACTED/local/lib/python2.7/site-packages/tensorflow/python/framework/ops.py", line 1298, in __init__
    self._traceback = _extract_stack()

InvalidArgumentError (see above for traceback): Cannot assign a device to node 'add': Could not satisfy explicit device specification '/device:GPU:0' because no devices matching that specification are registered in this process; available devices: /job:localhost/replica:0/task:0/cpu:0
	 [[Node: add = Add[T=DT_FLOAT, _device="/device:GPU:0"](Const, Const_1)]]

The main error now shows up at the end, which is where people are
likely to find it more easily.
Change: 133429936