docs: document strict value+type semantics of moonblade in/not in

[mvanhorn]

Summary

The in / not in operator docs now warn that membership uses strict equality matching both value and type, with the cast workaround (int(col) in [1, 2, 3]). The empty help strings for x in y and x not in y are filled in, and the generated Markdown docs are regenerated via ./scripts/docs.sh.

Why this matters

#1095: @eeweegh found that int(b) not in [1, 2, 3] stopped matching after moonblade switched contains/in to strict value-and-type equality. @Yomguithereal confirmed the strict semantics are intended and asked for a doc warning ("a warned person makes for two"). The warning is scoped to list membership: string containers still cast the needle and accept regexes, so the prelude describes the list-membership strict-typing behavior the issue is about rather than claiming every x in y form is strict.

Testing

Docs-only change; ./scripts/docs.sh regenerated cheatsheet.md and functions.md so the embedded and generated help stay consistent. No Rust behavior change.

Fixes #1095