[tf.data] Optimize the implementation of DeserializeSparseOp<Variant>.

The most expensive part of this kernel is the index construction. The optimized implementation builds the new index matrix at most once, rather than performing up to 3 passes (adding a leading dimension, `SparseTensor::Concat()` and `Reshape()`), and adds a specialized codepath for the common case of stacking together rank-1 SparseTensors.

PiperOrigin-RevId: 202669432