feat: Introduce resize_mode parameter to GenerateVideosConfig and…

[poemsforaphrodite]

Summary

This PR adds support for the resize_mode parameter to GenerateVideosConfig for Veo 3 image-to-video generation, resolving issue #1895.

Problem

The Google Vertex AI REST API supports a resizeMode parameter for image-to-video generation ("pad" or "crop"), but this functionality was missing from the Python SDK's GenerateVideosConfig class. Users attempting to use this parameter received Pydantic validation errors.

Fixes #1895

Solution

Added resize_mode parameter as an optional string field following the existing pattern used by similar parameters (aspect_ratio, resolution, person_generation):

• Added field to GenerateVideosConfig class
• Added field to GenerateVideosConfigDict TypedDict for dict-based initialization
• Supports two values: "pad" (API default) and "crop"
• Auto-converts to resizeMode via existing camelCase conversion

Changes

Type Definitions (google/genai/types.py)

• Added resize_mode: Optional[str] field to GenerateVideosConfig class (line 9903)
• Added corresponding field to GenerateVideosConfigDict TypedDict (line 9977)

Tests (google/genai/tests/models/test_generate_videos.py)

• Added parametrized test cases for both "pad" and "crop" modes
• Added integration test test_image_to_video_resize_mode_poll for full operation lifecycle
• Updated test_all_parameters_vertex to include resize_mode with other parameters