Detach fval in torch_minimize to avoid memory leak (#2529)

Summary:
## Motivation

This fixes #2526

### Have you read the [Contributing Guidelines on pull requests](https://github.com/pytorch/botorch/blob/main/CONTRIBUTING.md#pull-requests)?

Yes.

Pull Request resolved: #2529

Test Plan:
The tests are passing.

## Related PRs
N/A

Reviewed By: esantorella

Differential Revision: D62529029

Pulled By: Balandat

fbshipit-source-id: 50da389d8a01659e5b510ac27c25363fc25af207

botorch/optim/core.py

@@ -192,11 +192,11 @@ def torch_minimize(
         else {name: limits for name, limits in bounds.items() if name in parameters}
     )
     for step in range(1, step_limit + 1):
-        fval, _ = closure()
+        fval = closure()[0].detach()
         runtime = monotonic() - start_time
         result = OptimizationResult(
             step=step,
-            fval=fval.detach().cpu().item(),
+            fval=fval.cpu().item(),
             status=OptimizationStatus.RUNNING,
             runtime=runtime,
         )