[Fix] Fix transposed prior grid in RTMOHead.switch_to_deploy for non-square input

[bolun365]

Motivation

When deploying RTMO with a non-square input size (e.g. 256x160, W×H),
the exported model (ONNX / MNN / TensorRT) produces wrong detections: a single
person yields many scattered high-confidence "phantom" boxes that NMS cannot
merge, even though the PyTorch model runs correctly.

The root cause is in RTMOHead.switch_to_deploy. It builds the dummy feature
maps used to precompute flatten_priors as:

torch.rand(1, 1, input_size[0] // s, input_size[1] // s)  # (W//s, H//s)

input_size is in (W, H) order, but a real feature map has shape
(H//s, W//s). For square inputs (e.g. 256x256, the default RTMO configs)
the two are equal so the bug is invisible. For non-square inputs the prior grid
is transposed, so boxes are decoded at the wrong locations during export.
(The runtime predict path is unaffected because it derives feature-map sizes
from the actual cls_score.shape[2:].)

Modification

In mmpose/models/heads/hybrid_heads/rtmo_head.py, RTMOHead.switch_to_deploy,
swap the two index orders so the dummy feature map matches the real
(H//s, W//s) layout:

torch.rand(1, 1, input_size[1] // s, input_size[0] // s)  # (H//s, W//s)

BC-breaking (Optional)

No. The change is a no-op for square input_size (all existing official RTMO
configs use square inputs), and only corrects the behavior for non-square
inputs, which was previously broken.


Use cases (Optional)

Deploying RTMO with a non-square (landscape/portrait) input_size, e.g.
input_size=(256, 160), now exports an ONNX/MNN/TensorRT model whose decoded
boxes match the PyTorch results.


## Checklist

**Before PR**:

- [ ] I have read and followed the workflow indicated in the [CONTRIBUTING.md](https://github.com/open-mmlab/mmpose/blob/master/.github/CONTRIBUTING.md) to create this PR.
- [ ] Pre-commit or linting tools indicated in [CONTRIBUTING.md](https://github.com/open-mmlab/mmpose/blob/master/.github/CONTRIBUTING.md) are used to fix the potential lint issues.
- [ ] Bug fixes are covered by unit tests, the case that causes the bug should be added in the unit tests.
- [ ] New functionalities are covered by complete unit tests. If not, please add more unit tests to ensure correctness.
- [ ] The documentation has been modified accordingly, including docstring or example tutorials.

**After PR**:

- [ ] CLA has been signed and all committers have signed the CLA in this PR.