Add "Context-Aware Sentence Merging" for Multi-Block Subtitle Translation

When translating subtitles into languages with reverse grammatical structures (e.g., English SVO to Japanese/Korean/Mongolian SOV), translating individual timestamped blocks sequentially breaks sentence logic. The verb or object is often stranded in a separate block, resulting in inaccurate or unreadable text.

We need a feature that automatically merges consecutive subtitle blocks into full sentences before translating (or sending them to the translation engine), and then intelligently re-splits them back into timestamps.