VariationalGaussianProcess layers can't handle batch size of 1

VariationalGaussianProcess layers cause models to raise InvalidArgumentErrors when running with a batch size of 1. This occurs both in model training and inference.

Custom code for replication is not required, the VGP tutorial itself crashes when a batch size of 1 is used.

Error message:
InvalidArgumentError: Requires start <= limit when delta > 0: 0/-1 for 'sequential_1/variational_gaussian_process_1/sequential_1_variational_gaussian_process_1_sequential_1_variational_gaussian_process_1_transposesequential_1_variational_gaussian_process_1_Independentsequential_1_variational_gaussian_process_1_VariationalGaussianProcess/value/sequential_1_variational_gaussian_process_1_sequential_1_variational_gaussian_process_1_transposesequential_1_variational_gaussian_process_1_Independentsequential_1_variational_gaussian_process_1_VariationalGaussianProcess/sample/sequential_1_variational_gaussian_process_1_transpose/forward/range' (op: 'Range') with input shapes: [], [], [] and with computed input tensors: input[0] = <0>, input[1] = <-1>, input[2] = <1>.

Start with the VGP tutorial referenced in the report and run it with a batch size of 1 during both training and inference. Trace the VariationalGaussianProcess layer at the failing Range operation, then verify that the tutorial completes without the reported InvalidArgumentError for batch size 1.