[aotinductor] Remove output_spec from AOTInductorModelCache (#110462)
Summary: No need to store output_spec as the returned exported.call_spec already contains that.
Pull Request resolved: https://github.com/pytorch/pytorch/pull/110462
Approved by: https://github.com/angelayi