DeepSpeed: [BUG] "Setting ds_accelerator" uses `print` not `logger`

Describe the bug When accelerate tries to run any command while DeepSpeed is available, we get "Setting ds_accelerator ...", which while not explicitly bad, is making a test fail due to a non-clean CLI print when non-deepspeed code is used (we have a check for is_deepspeed_available() which is probably triggering this.

As a nice QOL, it would be good to have get_accelerator (https://github.com/microsoft/DeepSpeed/blob/master/accelerator/real_accelerator.py#L102) and set_accelerator (https://github.com/microsoft/DeepSpeed/blob/master/accelerator/real_accelerator.py#L109) use the already-existing logger in the framework if possible, so that we can disable these annoying prints when they’re not needed 😃

To Reproduce Steps to reproduce the behavior:

1. pip install accelerate deepspeed -U
2. Create a file with:

from accelerate.commands.tpu import tpu_command_parser, tpu_command_launcher

parser = tpu_command_parser()
args = parser.parse_args([
    "--config_file", "tests/test_configs/latest.yaml",
    "--install_accelerate",
    "--debug"
])

tpu_command_launcher(args)

1. Run python {my_file_name.py}
2. Should print:

Setting ds_accelerator to cuda (auto detect)
Running gcloud compute tpus tpu-vm ssh test-tpu --zone us-central1-a --command cd /usr/share; pip install accelerate -U; echo "hello world"; echo "this is a second command" --worker all

Expected behavior

A configurable option to silence these print statements by having them run through the logging system instead, perhaps as an info or as a debug.