Fix some issues in official tf.nn.topk() in lite

[scottcjt]

Not sure if Google is planning a breaking change. If not:

1. The topk_v2 implementation has its output values and indices in the opposite order.

   Below is TensorFlow Python API behavior:

   In [5]: x
   Out[5]: <tf.Tensor 'x:0' shape=(100,) dtype=float32>
   
   In [7]: y, z = tf.nn.top_k(x, k=10)
   
   In [8]: y
   Out[8]: <tf.Tensor 'TopKV2:0' shape=(10,) dtype=float32>
   
   In [10]: z
   Out[10]: <tf.Tensor 'TopKV2:1' shape=(10,) dtype=int32>

2. Output array types are not correctly propagated.

3. First output array name should not be appended with ":0" because it is deliberately removed here.

4. If k is provided at run time, two output arrays will be dynamic, and have shape [..., 0] here. I changed toco to only check constant buffers instead, RFC.