`torch.cat`/`torch.stack` for static Sequential cases

Docs: https://docs.pytorch.org/docs/stable/generated/torch.cat.html

Initial implementation should only tackle statically defined sequences (lists, tuples) at the call-site.

Tracking list comprehensions or variables mapping to a sequence is out-of-scope.

### Example

```python
def concat_emb(self, c3: F[T, "B F H3 W3"], embeddings: F[T, "B E"]) -> F[T, "B F H3*W3+E"]:
"""Concat c3 with embeddings."""
_, F, h_3, w_3 = c3.shape
# expand to concat: [B, E] -> [B, F, E]
embeddings = embeddings.unsqueeze(1).expand(-1, F, -1)
# this can be known statically since embeddings and c3 are known
c3_flat = torch.cat([c3.flatten(-2), embeddings], dim=-1)
return c3_flat
```

Start with the linked torch.cat documentation and the repository’s existing handling of torch calls. Implement support for statically defined lists and tuples at the call site for torch.cat and torch.stack; leave list comprehensions and variables that refer to sequences out of scope, and verify the provided example’s shapes are inferred.

Written by the indexing model from the issue text.