WAN2.1 FLF2V: Incorrect MASK Creation????

Hello! I think that it is maybe error. (Or not, please explain it for me!!)

In **WanImageToVideoPipeline** class in `pipline_wan_i2v.py`,

(the code is the part of `prepare_latents` function)

**For I2V**, masking shape like as below:
```
[[1, 0, 0, ... , 0]
[1, 0, 0, ... , 0]
[1, 0, 0, ... , 0]
[1, 0, 0, ... , 0]]
```
I understood: when the mask is 1, input video frame does not change.
(*Mask shape: [1, 4, 21, 60, 104] = [B, C, F, H, W])

**But in the FLF2V case,** masking shape like as below:
```
[[1, 0, 0, ... , 0]
[1, 0, 0, ... , 0]
[1, 0, 0, ... , 0]
**[1, 0, 0, ... , 1]]**
```
Here, **why the last frame mask has 1 only in last channel??**
Is there anyone who can explain this part?

Start in pipline_wan_i2v.py at the WanImageToVideoPipeline.prepare_latents function and compare the mask construction for I2V with the FLF2V case. Trace how the [B, C, F, H, W] mask is formed and determine whether the final-channel value is intentional or an error. Done means documenting the reason or identifying the precise behavior that needs correction.

Scritto dal modello di indicizzazione a partire dal testo della issue.