Add JSON initialization methods about num.cr HOT 2 CLOSED

There isn't even a good way to convert a Tensor to Array right now. I need to work on some serializers, I'll probably get around to this when I actually implement reading/writing from files.

from num.cr.

Closed by #23

I handle the one dimensional case only, but you can pass a shape, so to store N-dimensional just store the shape separately.

json_str = "[1, 2, 3]"

puts Tensor(Int32).from_json(json_str, [3])
# [1, 2, 3]

puts Tensor.random(0...10, [3]).to_json
# [7,6,2]

from num.cr.