enter/exit_autocast of torch.amp.autocast_mode

Nobody has claimed this yet.

## 🚀 Model / language coverage

I'm trying to get a fuller picture of what we need to support NeVA. As such I'm using:

```python
def thunder_backend(gm, args):
gm.real_recompile()
from thunder.examine import examine
try: # Examine may raise an exception
thunder.examine.examine(gm, *args)
except Exception as e:
print(f"Hit problem with examine:\n{e}")
# Don't really use Thunder just return the original graph
return gm

...
#model.model = thunder.jit(model.model)
model.model = torch.compile(backend=thunder_backend)(model.model)
```
(thanks Ivan for the great idea!)

And one of the issues that gets reported is e.g.

```
Found 2 distinct operations, of which 0 (0.0%) are supported
Please file an issue requesting the following operators here: https://github.com/Lightning-AI/lightning-thunder/issues/
new
_enter_autocast of torch.amp.autocast_mode
_exit_autocast of torch.amp.autocast_mode
```

### Pitch

This looks like it is going to be important for #343.

### Alternatives / Potential work-arounds

### Minimal Repro

cc @apaz-cli @crcrpar @tfogal

Start by reproducing the reported torch.compile case with the thunder_backend example and inspect how torch.amp.autocast_mode operations are handled. The work is complete when _enter_autocast and _exit_autocast are supported and no longer appear in the unsupported-operations report.

Written by the indexing model from the issue text.