[python] Support encoding images as video

[YannByron]

Purpose

Allow PyPaimon users to append ordered image bytes or Pillow images to a video-frame field without first creating an MP4 themselves. MultimodalTable.add_images_as_video encodes the images with explicit FPS, codec, pixel format, GOP size, and codec options, then reuses the existing add_video descriptor and blob write path.

The encoder validates non-empty input, equal image dimensions, and a one-to-one match between images and table rows. Encoding completes in a temporary file before a Paimon commit, and partial files are removed on failure. The optional pypaimon[video] extra supplies PyAV and Pillow on Python 3.10 or newer.

Tests

• Encode PNG bytes into MP4 and decode the resulting frames with PyAV.
• Reject invalid encoding parameters, mixed dimensions, and image/row count mismatches.
• Verify video-frame descriptors and stored payloads through the table API.
• Run the affected multimodal test suites: 110 passed, 5 skipped, 24 subtests passed.
• Run the Python flake8 configuration and Apache RAT license scan.

[Akash3121]

LGTM

The image-to-video encoder validates inputs, streams image iteration into a temporary MP4, cleans partial output on encoding failures, verifies image/row cardinality before writing, and safely reuses the existing descriptor-backed video path. The optional dependency markers preserve Python 3.6/3.7 compatibility, while CI exercises the new extra on Python 3.10.