fix(wan_i2v): repeat conditioning images along batch dimension in prepare_latents

- Repeat image and last_image along batch dimension in WanPipelineI2V_2_1 and WanPipelineI2V_2_2 to match effective_batch_size, resolving concatenation dimension mismatch when batch_size > 1.
- Validate divisibility of batch_size against image and last_image batch sizes.
- Validate alignment between image and last_image batch sizes when both are greater than 1.
- Add focused unit tests in wan_i2v_prepare_latents_test.py.