[enhancement]: Add video duration in seconds with automatic frame count calculation

### What should this feature add?

Add a **Duration (seconds)** field to the video generation workflow forms, including the built-in Wan 2.2 Image-to-Video Lightning workflow. Calculate a supported frame count automatically from the requested duration and FPS.

Currently, the workflow exposes Frames and FPS. To request a six-second clip at 16 FPS, users need to calculate the frame count themselves and account for Wan's supported frame counts (4n + 1), resulting in 97 frames. A duration field would make this much easier to understand.

### Suggested behavior

- Let users enter the desired duration in seconds, such as 6 or 10.
- Convert the duration and FPS to the nearest frame count supported by the selected model, respecting its limits.
- Show the resulting frame count and actual duration when rounding is required.
- Keep Frames available as an advanced control for users who need it.
- Keep duration and frame count consistent when FPS changes.

For example, entering 6 seconds at 16 FPS in a Wan workflow could select 97 frames and display approximately 6.06 seconds.

### Alternatives

A calculated duration readout beside Frames would help, but entering seconds directly would be more intuitive.

### Additional context

Using InvokeAI 6.14.0 and the built-in Image to Video - Wan 2.2 Lightning workflow.

I searched open and closed issues for duration, video length, frames, and Wan. #8903 mentions duration for an API-driven remote video generation node; this request is for duration controls in local video workflow forms.

No files, tests, or entry points are named. Start by locating the video generation workflow forms and the built-in Wan 2.2 Image-to-Video Lightning workflow, then trace how Frames, FPS, model limits, and supported frame counts are represented. Done means duration input, automatic supported-frame rounding, actual-duration feedback, and consistent updates when FPS changes.

Written by the indexing model from the issue text.