[PLAN] Improve performance with dimension compaction and indexer

[sonots]

1. Stop using ndloop and compute an operation with one CUDA kernel using indexer
2. Compact dimension to make computation of indexer fast

Element-wise (binary ops) is already done at https://github.com/sonots/cumo/pull/64. But, reduction and others such as store_from are not yet done.

Without this, cumo (and red-chainer) can not compete with cupy (and chainer)

Current performance comparison on k80 machine:

• chainer mnist: 5 sec / epoch
• red-chainer mnist: 13 sec / epoch