Allow to retain and moderate unsure posts instead of showing CAPTCHA

Goal

• When text analysis is unsure, retain posts as "unpublished" for manual moderation instead of showing a CAPTCHA.

Details

• The default behavior for an "unsure" response is to show a CAPTCHA.
• The Mollom module already supports to retain posts for manual moderation when text analysis identifies spam.
• What is missing and desired is to retain posts for manual moderation when text analysis is unsure, instead of showing a CAPTCHA.

Related issues

• #881534: Allow to unpublish spam posts instead of rejecting — added fundamental support for retaining posts as unpublished.
• #1096412: Allow to disable fall-back CAPTCHA for text analysis — considers to add support for Mollom's "binary mode", which makes Mollom only return "spam" or "ham", but never "unsure." That is not helpful for this issue though, since we want to react on "unsure". Thus, telling the backend to never return it would be incompatible.
• #364575: Support alternative spam rejection methods than a CAPTCHA (e.g. unpublishing) — intends to introduce a more sophisticated event system for Mollom module, allowing to precisely specify the actions to perform for various triggers.