Fix Wishart distribution documentation (#95816)
This PR fixes the `torch.distributions.wishart.Wishart` example.
Running the current example
```python
m = Wishart(torch.eye(2), torch.Tensor([2]))
m.sample() # Wishart distributed with mean=`df * I` and
# variance(x_ij)=`df` for i != j and variance(x_ij)=`2 * df` for i == j
```
fails with
```
---------------------------------------------------------------------------
ValueError Traceback (most recent call last)
Untitled-1 in
[321](untitled:Untitled-1?line=320) # %%
----> [322](untitled:Untitled-1?line=321) m = Wishart(torch.eye(2), torch.Tensor([2]))
[323](untitled:Untitled-1?line=322) m.sample() # Wishart distributed with mean=`df * I` and
[324](untitled:Untitled-1?line=323) # variance(x_ij)=`df` for i != j and variance(x_ij)=`2 * df` for i == j
Untitled-1 in __init__(self, df, covariance_matrix, precision_matrix, scale_tril, validate_args)
[83](untitled:Untitled-1?line=82)
[84](untitled:Untitled-1?line=83) if param.dim() < 2:
---> [85](untitled:Untitled-1?line=84) raise ValueError("scale_tril must be at least two-dimensional, with optional leading batch dimensions")
[86](untitled:Untitled-1?line=85)
[87](untitled:Untitled-1?line=86) if isinstance(df, Number):
ValueError: scale_tril must be at least two-dimensional, with optional leading batch dimensions
```
Is seems that the parameters of `Wishart.__init__()` were re-ordered, but the documentation was not updated.
This PR fixes it. Here is the updated behaviour:
```python
m = Wishart(torch.Tensor([2]), covariance_matrix=torch.eye(2))
m.sample()
```
```
Untitled-1:255: UserWarning: Singular sample detected.
tensor([[[6.6366, 0.7796],
[0.7796, 0.2136]]])
```
Pull Request resolved: https://github.com/pytorch/pytorch/pull/95816
Approved by: https://github.com/ngimel, https://github.com/kit1980
