Type checking on sample for num_samples

[sabinala]

Need several tests for num_samples:

• num_samples = 3 as an int
• num_samples = 3.0 as a 'float`
• num_samples = 3.1 as a 'float`
• num_samples = torch.tensor(3) as tensor
• num_samples = torch.tensor([3]) as tensor
• num_samples = torch.tensor(3.0) as tensor
• num_samples = torch.tensor([3.0]) as tensor
• num_samples = None
• num_samples = -3
• num_samples = "a"
• num_samples = [3, 4, 5] as an array

We check that the output either "fails gracefully" with an informative error message, or warns us that it has converted to an int.

[sabinala]

Simplest solution: throw an informative error if input to num_samples is not a positive int.