Deprecate trainer.num_processe/trainer.num_gpus and remove incorrect tests

[four4fish]

Proposed refactor

Trainer.num_processes() and trainer.num_gpus are not used in the code base and only exists in tests. Propose deprecate/remove these properties and remove incorrect/unnecessary tests

Motivation

1. Simplify code and reduce confusion. Strategy.num_processes != Trainer.num_processes, and Trainer.num_processes is only called in test. (confusion raised in Lazy initialize Strategy.parallel_devices #11572)
2. Simplify accelerator_connector rewrite Rewrite Accelerator_connector and follow up tasks #11449. The current accelerator_connector has a lot of logic related to num_processes which is not necessary and confusing, remove it first will simplify the refactor

Pitch

Steps:
1 Deprecate trainer.num_processes
https://github.com/PyTorchLightning/pytorch-lightning/blob/fe34bf2a653ebd50e6a3a00be829e3611f820c3c/pytorch_lightning/trainer/trainer.py#L1969-L1971

2. Remove trainer.num_processes tests
   https://github.com/PyTorchLightning/pytorch-lightning/blob/fe34bf2a653ebd50e6a3a00be829e3611f820c3c/tests/trainer/test_trainer.py#L2093-L2111

3. Do not carry self.num_processes logic over to accelerator_connector rewrite.

Notes: this won't impact Strategy.num_processes as it's irrelevant. Strategy.num_processes calculated base on parallel_devices which is not equal or related to trainer.num_processes

same for trainer.num_gpus

Additional context

---

If you enjoy Lightning, check out our other projects! ⚡

• Metrics: Machine learning metrics for distributed, scalable PyTorch applications.

• Lite: enables pure PyTorch users to scale their existing code on any kind of device while retaining full control over their own loops and optimization logic.

• Flash: The fastest way to get a Lightning baseline! A collection of tasks for fast prototyping, baselining, fine-tuning, and solving problems with deep learning.

• Bolts: Pretrained SOTA Deep Learning models, callbacks, and more for research and production with PyTorch Lightning and PyTorch.

• Lightning Transformers: Flexible interface for high-performance research using SOTA Transformers leveraging Pytorch Lightning, Transformers, and Hydra.

cc @justusschock @awaelchli @akihironitta @rohitgr7 @kaushikb11 @Borda @ananthsub @ninginthecloud @jjenniferdai

[awaelchli]

This is reasonable imo considering the genericTrainer(devices=x) argument was recently introduced., num_processes does not follow that terminology anyway. When deprecating num_processes, num_gpus etc., we should introduce num_devices imo.
Other than that, I think we can go ahead with this change.

[carmocca]

Is this finished?

[rohitgr7]

looks like the deprecation is done already. Closing this.