Add half_t support for batch_dot.

[eric-haibin-lin]

apache/mxnet#11796
@DickJC123 @szhengac @szha

crude benchmark

import mxnet as mx
a = mx.nd.ones((100,100,100), ctx=mx.gpu(), dtype='float16')
b = mx.nd.ones((100,100,100), ctx=mx.gpu(), dtype='float16')
for i in range(10):
    c = mx.nd.batch_dot(a,b)
mx.nd.waitall()
import time
begin = time.time()
for i in range(500):
    c = mx.nd.batch_dot(a,b)
mx.nd.waitall()
end = time.time()
print(end - begin)

Before:
python tests.py
0.9715321064

After:
python tests.py
0.0328528881073

Passed existing gpu batch_dot unit test in MXNet locally.

[szha]

awesome!