easterbrooka teaches Mike to actually read Outlook email

Outlook composes most mail as HTML, which meant the plain-text body slot the extractor was reading came back empty - and the fork was quietly handing the AI a blank message. easterbrooka's fix falls back to the HTML body when the plain-text version is missing, matching how the fork already handles standard email files.

The second bug hit forwarded chains: when someone attaches an email to another email, the underlying parser library labels those differently from regular file attachments, and the fork was filtering them out. They now flow through to both the user-facing attachment list and the AI's recursive read of the thread. Six new tests lock the behaviour in.