Tidy-up feasibility evaluation in optimize_mixed_alternating (#2952)

Summary:
<!--
Thank you for sending the PR! We appreciate you spending the time to make BoTorch better.

Help us understand your motivation by explaining why you decided to make this change.

You can learn more about contributing to BoTorch here: https://github.com/pytorch/botorch/blob/main/CONTRIBUTING.md
-->

## Motivation

As discussed with esantorella in #2944 (comment), this PR tidies-up the feasibility evaluation in `optimize_mixed_alternating.` It is now refactored in a way that it uses `optim.parameter_constraints.evaluate_feasibility`, which improves code sharing.

Regarding nonlinear inequality constraints that were also discussed in the thread above, I faced the problem that nonlinear constraints only work with a batch limit of 1, which would render `optimize_mixed_alternating` highly inefficient as it makes use of batch optimization. So, I think, we have to first get rid of the limitation of `batch_limit==1` for nonlinear constraints. Should I create a separate issue for this?

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

Yes.

Pull Request resolved: #2952

Test Plan: Unit tests.

Reviewed By: Balandat

Differential Revision: D79644900

Pulled By: saitcakmak

fbshipit-source-id: d7bfa0905ffc489f79d815f97a4444f5914e8915

Author: jduerholt

botorch/optim/optimize_mixed.py

@@ -22,6 +22,7 @@
     _validate_sequential_inputs,
     OptimizeAcqfInputs,
 )
+from botorch.optim.parameter_constraints import evaluate_feasibility
 from botorch.optim.utils.acquisition_utils import fix_features, get_X_baseline
 from botorch.utils.sampling import (
     draw_sobol_samples,
@@ -134,17 +135,14 @@ def _filter_infeasible(
     Returns:
         The tensor `X` with infeasible points removed.
     """
-    if inequality_constraints is None and equality_constraints is None:
-        return X
-    is_feasible = torch.ones(X.shape[:-1], device=X.device, dtype=torch.bool)
-    if inequality_constraints is not None:
-        for idx, coef, rhs in inequality_constraints:
-            is_feasible &= (X[..., idx] * coef).sum(dim=-1) >= rhs
-    if equality_constraints is not None:
-        for idx, coef, rhs in equality_constraints:
-            is_feasible &= torch.isclose(
-                (X[..., idx] * coef).sum(dim=-1), torch.tensor(rhs).to(coef)
-            )
+    # X is reshaped to [n, 1, d] in order to be able to apply
+    # `evaluate_feasibility` which operates on the batch level
+    is_feasible = evaluate_feasibility(
+        X=X.unsqueeze(-2),
+        inequality_constraints=inequality_constraints,
+        equality_constraints=equality_constraints,
+        nonlinear_inequality_constraints=None,
+    )
     return X[is_feasible]