[rl] bypass reload_weights by manually copy weights per-param, fix weight tying

[wwwjn]

Stack from ghstack (oldest at bottom):

• [WIP][rl] enable batch-invariant mode in RL loop #2395
• [rl] Generator enables TP, using torchtitan as Trainer, add grader for reward calculation #2244
• [rl] refactor save and load model weights using DCP #2221
• [rl] refactor torchtitan model registery in vllm #2194
• [rl] Use torchtitan config system for inference and simple GRPO #2191
• -> [rl] bypass reload_weights by manually copy weights per-param, fix weight tying #2410

This is a alternative fix to #2402 (comment).

Weight updating between trainer and generator is totally broken because: It's caused by we called "reload_weights" when updating the weights. The reload_weights has following steps:

• initialize_layerwise_reload(model): Saves the current real GPU tensors as info.kernel_tensors, and replace all parameters with meta tensor.
• Call model.load_weights(weights_iter): This function is written by us and calls set_model_state_dict, Internally, set_model_state_dict tries to do param.data.copy_(loaded_weight) for each parameter. When parameters are meta tensor, it will do "no-op". So the weights never get updated

In this PR:

• Totally bypass reload_weights, and don't load from a file when we update the weights
• Gets the model via self.engine.model_executor.driver_worker.get_model()
• Iterates over model.named_parameters() to find the matching parameter by name
• Does param.data.copy_(new_tensor) directly