Teams chat/channel search tools don't expose read/unread status

### Problem Statement

I'm trying to find **unread Teams messages/channels** via the **Microsoft 365 connector**, but `chat_message_search` and `teams_list_chats` currently return no read/unread indicator at all (no `isRead`, no `lastMessageReadDateTime`, etc.) - even though Microsoft Graph exposes per-member read state (`chat.lastMessageReadDateTime`) that could be used to compute this.

As a result, there is no way to ask Claude "what unread Teams messages do I have" without falling back to keyword/date-range search and manually eyeballing what has already been seen.

By comparison, `outlook_email_search` already returns an `isRead` field for emails, so this looks like an inconsistency between the connector's email and Teams read-state handling rather than an underlying Graph API limitation.

### Proposed Solution

Add a read/unread indicator to the Teams tools' output - e.g. a boolean derived from comparing each message's timestamp against the chat member's `lastMessageReadDateTime`, mirroring how `outlook_email_search` already exposes `isRead` for email.

It would also help to add an optional filter parameter (e.g. `unreadOnly`) on `chat_message_search` and `teams_list_chats`, so unread state can be queried directly instead of inferred from date ranges.

### Alternatives Considered

Falling back to `outlook_email_search` and its `isRead`-style filtering isn't possible for Teams today - the only workaround is searching `chat_message_search` by keyword/date range and manually checking what's already been seen, which doesn't scale across many chats and channels.

### Area

Tool Discovery / Invocation

### Priority

Medium - Would be helpful

### Additional Context

_No response_

Begin with the chat_message_search and teams_list_chats entry points, then compare their response fields with outlook_email_search and Microsoft Graph's chat.lastMessageReadDateTime. Done means Teams results expose a defined read/unread value and, if accepted, an unreadOnly filter behaves consistently for both tools.

Vom Indexierungsmodell aus dem Issue-Text verfasst.