GetConvolve*Algorithms return tensor-op algos

An additional loop to toggle tensor_ops is not needed when enumerating
convolution algorithms. Also tensor_ops are not included in the returned list
if they are not supported by the cuDNN or GPU architecture versions.