Add constraint to not partition standalone batch norm (#1501)

Summary:
Pull Request resolved: https://github.com/pytorch/executorch/pull/1501

The XNNPACK backend does not current support lowering standalone (non-fused) batch norms. This will be done in the near future, but models with standalone batch norms are failing to lower as the op is partitioned but cannot be lowered.

This change adds a op-level constraint for batch norm to the XNNPACK partitioner to only partition batch norms that can be fused. This constraint will be relaxed once standalone batch norm is fully supported.

Reviewed By: mcr229

Differential Revision: D52491544

fbshipit-source-id: 861744d836b0cbfc07700bc411e9677ba80367df