Bring back TIMM model inductor CI test (#87730)
Summary: https://github.com/pytorch/pytorch/pull/87588 has solved the
inductor compilation speed regression, so we can try to run TIMM models
with fewer shards and also enable pretained model downloading which
should resolve the flakyness we have seen previously.
cc @jansel @mlazos @soumith @voznesenskym @yanboliang @penguinwu @anijain2305
Pull Request resolved: https://github.com/pytorch/pytorch/pull/87730
Approved by: https://github.com/anijain2305
