automated: linux: torchbench: fix git submodules and iterations argument

Fix two issues with TorchBench execution:

1. Add git submodule initialization after cloning the repository.
   TorchBench uses submodules that need to be initialized before
   installing models.

2. Change --iterations to -n for the run.py command. TorchBench's
   run.py uses -n for the number of iterations, not --iterations.
   This fixes the AssertionError about unknown args.

Signed-off-by: Anders Roxell <anders.roxell@linaro.org>
