Making protected params of MobileNetV3 public

[datumbox]

This PR makes the previously protected constructor parameters of MobileNetV3 public.

Back when the model was added (see #3276 (comment)), it was unclear whether any models would use those parameters. Now with segmentation and detection models in place using them, we should properly expose them to the users in case they want to create their own backbones.

Along with unit-tests, I confirmed that all current MobileNetV3 models remain unaffected by running:

cd ./references/detection
python -m torch.distributed.launch --nproc_per_node=2 --use_env train.py --model ssdlite320_mobilenet_v3_large --test-only --pretrained
python -m torch.distributed.launch --nproc_per_node=2 --use_env train.py --model fasterrcnn_mobilenet_v3_large_fpn --test-only --pretrained
python -m torch.distributed.launch --nproc_per_node=2 --use_env train.py --model fasterrcnn_mobilenet_v3_large_320_fpn --test-only --pretrained

cd ./references/segmentation
python -m torch.distributed.launch --nproc_per_node=2 --use_env train.py --model lraspp_mobilenet_v3_large --test-only --pretrained
python -m torch.distributed.launch --nproc_per_node=2 --use_env train.py --model deeplabv3_mobilenet_v3_large --test-only --pretrained

cd ./references/classification
python -m torch.distributed.launch --nproc_per_node=2 --use_env train.py --model mobilenet_v3_large --test-only --pretrained
python -m torch.distributed.launch --nproc_per_node=2 --use_env train.py --model mobilenet_v3_small --test-only --pretrained
python -u train_quantization.py --device='cpu' --model='mobilenet_v3_large' --test-only