Implement batch_det function

[takuseno]

Hi, @TE-AkioHayakawa san, @TE-TakuyaNarihira san.

I've implemented batch_det function that computes determinant of input array.

a = nn.Variable((2, 13, 13))
det = F.batch_det(a) # det.shape == (2,)

nd = np.random.random((2, 13, 13))
a.d = nd
det.forward()

assert np.allclose(det.d, np.array(list(map(np.linalg.det, nd))))

batch_det works with forward and backward correctly now. cuda extension is implemented.

[takuseno]

cuda version is implemented sony/nnabla-ext-cuda#132

[takuseno]

@TE-KazukiYoshiyama san reviewed this

[takuseno]

@TE-AkioHayakawa
I've resolved conflict. Now, it's ready for merge :)

[AkioHayakawa-sony]

@takuseno
Thanks a lot!
Could you do the same thing for sony/nnabla-ext-cuda#132 ?

[takuseno]

@TE-AkioHayakawa
I've resolved that conflict. The cuda implementation is ready for merge too :)

[AkioHayakawa-sony]

@takuseno
One more minor thing. Could you update doc/python/api/function.rst?
FYI, see https://github.com/sony/nnabla/pull/550/files as a reference.

[takuseno]

@TE-AkioHayakawa
I've added batch_det to document. And, squashed version of PR is created as #553 . If you prefer the new PR, I'd like you to see it :)

[AkioHayakawa-sony]

@takuseno
Thanks a lot! It looks Nice!
I will see #533 and run CI testing again. Please wait a moment.