Fix documentation issue with `tf.nn.conv1d`

The `tf.nn.conv1d` supports float16, float32, and float64
though in `tf.nn.conv1d.__doc__` only float16 and float32
are mentioned. This fix updates the doc string to add
float64 as the supported data type.

Signed-off-by: Yong Tang <yong.tang.github@outlook.com>