find_unused_parameters forced to False for ddp_{spawn,fork}

[LoicGrobol]

🐛 Bug

Since #13744 , the ddp_fork and ddp_spawn strategies have find_unused_parameters set to False, making them equivalent to ddp_fork_find_unused_parameters_false and ddp_spawn_find_unused_parameters_false. This is caused by a mistaken simplification of register_strategies that removed the distinction between them. Besides being an API break, this can trigger a runtime error when training any model with unused parameters using these strategies

To Reproduce

https://colab.research.google.com/drive/1xxoSnEDc8s6z-j66o1MbtEjVwGJ7K43p

Environment

• CUDA:
  • GPU:
    • Tesla T4
  • available: True
  • version: 11.3
• Packages:
  • lightning: None
  • lightning_app: None
  • numpy: 1.21.6
  • pyTorch_debug: False
  • pyTorch_version: 1.12.0+cu113
  • pytorch-lightning: 1.7.0
  • tqdm: 4.64.0
• System:
  • OS: Linux
  • architecture:
    • 64bit
  • processor: x86_64
  • python: 3.7.13
  • version: 1 SMP Sun Apr 24 10:03:06 PDT 2022

Please tell me if I can just submit a PR (I can probably have it ready today) or if you'd rather have a go at this yourselves.

cc @tchaton @rohitgr7 @justusschock @kaushikb11 @awaelchli @akihironitta