feat(scan): guided engine pick + examples in --help

[Maksim-Gr]

Addresses two user-friendliness gaps: the two-step scan → table workflow, and poor discoverability without the README.

What changed

Guided generation in scan. After printing suggestions, scan now offers to generate a migration from one of them, so you don't have to re-type a full table --engine X --order-by ... command:

Pick an engine to generate [1-3, Enter to skip]: 2
Written: ./video_events_up.sql
Written: ./video_events_down.sql

• Only prompts when stdin is a real terminal and the input is a file (not -). Piped/scripted/redirected runs are byte-for-byte unchanged — no prompt, no files.
• Empty / invalid / out-of-range input skips silently (today's behavior).
• Writes {name}_up.sql/{name}_down.sql to the current dir using the picked suggestion's engine + ORDER BY. scan -n now names those files.

Examples in --help. Added an EXAMPLES section to the top-level command and every subcommand (scan/table/kafka/diff), reusing the README's example invocations.

Demo GIF. Added a VHS terminal demo (assets/demo.gif + assets/demo.tape, regenerable with vhs assets/demo.tape) and embedded it at the top of the README.

Refactor. Extracted the table-generation body into generate_table_migration(), shared by the table command and the new scan prompt (no behavior change for table).