Optimizers.scheduler_step() bug fix

[JADGardner]

I believe there is a bug in the Optimizers.scheduler_step() method:

Calling Optimizers.scheduler_step() results in the error:

'dict' object has no attribute 'param_group_name'

Here is a minimal script to produce the error:

import torch
from nerfstudio.engine.optimizers import Optimizers, AdamOptimizerConfig
from nerfstudio.engine.schedulers import ExponentialDecaySchedulerConfig

optimizers_config = {
    "fields": {
        "optimizer": AdamOptimizerConfig(lr=1e-2, eps=1e-15),
        "scheduler": ExponentialDecaySchedulerConfig(lr_final=0.0001, max_steps=200000),
    }
}

dummy_params = [torch.tensor([42.0])]
param_groups = {"fields": dummy_params}

optimizer = Optimizers(optimizers_config, param_groups)
optimizer.optimizer_step("fields")
optimizer.scheduler_step("fields")

[tancik]

Thanks for the fix!