jit: `torch.cuda.stream` and other related functionality are silently ignored when jitting.

Nobody has claimed this yet.

As found by @syed-ahmed, `thunder.jit` silently ignores the use of `torch.cuda.stream` in the jitted function. There should a warning or an error. As suggested by @IvanYashchuk, we could also make this a SharpEdge.

```python
import torch
import thunder

def func(a):
s = torch.cuda.Stream()
with torch.cuda.stream(s):
for i in range(3):
a = torch.matmul(a, a)
a = torch.nn.functional.relu(a)
return a

a = torch.randn(256, 256, device=torch.device("cuda"))

# Thunder: ignores streams
jfunc = thunder.jit(func, sharp_edges=thunder.SHARP_EDGES_OPTIONS.ERROR)
jfunc(a)

# Eager: uses streams
func(a)
```

cc @apaz-cli

Start with the reproduction in the issue and compare eager execution with thunder.jit when using torch.cuda.Stream and torch.cuda.stream. Trace how sharp_edges=thunder.SHARP_EDGES_OPTIONS.ERROR handles unsupported functionality. Done means stream usage is no longer silently ignored and the chosen warning, error, or SharpEdge behavior is covered by a regression test.

Written by the indexing model from the issue text.