Bug: AttributeError in infer() when use_fov_head=False and f_px not provided

[McVyp]

Description

When creating a model with use_fov_head=False and calling infer() without
providing f_px, the code crashes with an unhelpful AttributeError.

Steps to Reproduce

from depth_pro.depth_pro import DepthProConfig, create_model_and_transforms
import torch

config = DepthProConfig(
    patch_encoder_preset="dinov2l16_384",
    image_encoder_preset="dinov2l16_384",
    decoder_features=256,
    use_fov_head=False,
    checkpoint_uri=None,
)
model, transform = create_model_and_transforms(config=config)
model.eval()

image = torch.rand(1, 3, 256, 256)
model.infer(image)

Error

Traceback (most recent call last):
  File "<stdin>", line 26, in <module>
  File "<stdin>", line 23, in infer
AttributeError: 'NoneType' object has no attribute 'to'

Root Cause

When use_fov_head=False, forward() returns fov_deg=None.
The code unconditionally calls fov_deg.to() with no None guard.

Expected Behaviour

A clear ValueError telling the user that f_px must be provided
when use_fov_head=False.