…ion in MCP server tests

…crash

Add an ErrorBoundary and wrap each answer item in it, so a failure in one
section degrades only that section instead of blanking the whole message.
Also guard the optional streamable chain in StandardResourceCard that could
throw on a null value and take down the message.

)

The Expert can ask 1-4 clarifying questions in a single turn, each rendered as
its own single- or multi-select option card; all answers are collected before
the turn is submitted. Answered cards can be edited and resubmitted, and a card
from a past turn is disabled once a newer message arrives.

Adds a follow-up-questions cadence setting (all at once vs one at a time) in the
composer settings menu, shipped to the agent via the expert context.

Add an always-visible Plan mode toggle to the composer. When enabled, the Expert
proposes a plan instead of making changes, rendered as a plan card with Approve,
Edit, Request changes and Reject actions:

- Approve exits plan mode and proceeds with the plan.
- Edit loads the plan markdown into the composer for direct editing.
- Request changes focuses an empty composer to describe a change in words.
- Reject abandons the plan.

The plan card renders its markdown through RichContent (passing the message and
answer uuids it requires), and reuses the composer's pending-input and auto-grow
behaviour. Plan mode and the approval signal are shipped to the agent via the
expert context.

Plan mode is only meaningful inside the instance/device editor for now,
so gate the composer toggle on immersive mode and force the persisted
planMode off whenever the user is outside immersive (including on load),
preventing a stale value from being sent in non-immersive contexts.

- Guard the optional streamable chain in FlowResourceCard directly instead
  of relying on a render boundary to mask the throw. Reduce ErrorBoundary to
  a single last-resort backstop per answer item in AiMessage; drop the
  per-section boundary wrappers in AnswerWrapper.
- Rewrite QuestionsList on top of the existing ff-radio-group (single-select)
  and ff-checkbox (multi-select) components so options look like standard,
  clickable form controls and stay consistent with the rest of the app.
- Replace the imperative growComposerToContent DOM measuring with CSS
  field-sizing on the textarea; drop the manual reflows and the auto-grown
  flag. The composer auto-sizes to content and pins to an explicit height
  only after a drag-resize.

Per review: the local-catch pattern was the only one in the frontend; the
rest of the app leans on the global app.config.errorHandler. The real throws
are now guarded at their source (the optional streamable chains in the
resource cards), so the boundary was redundant. Remove it entirely and let
genuinely unexpected render errors surface through the global handler like
everywhere else.

Replace the composer kebab menu with a settings gear that opens an
ff-dialog. The follow-up-questions cadence control now lives in the
dialog as an ff-radio-group, with a FormHeading per section so the
panel can grow as more settings are added.

…n-mode

# Conflicts:
#	frontend/src/components/expert/components/ExpertChatInput.vue
#	frontend/src/components/expert/components/messages/components/AnswerWrapper.vue

Add per-tool approval for the Expert's flow-building tools in the immersive
editor. The agent gates each tool call at the toolsNode seam by class
(read/write/delete) and per-tool preference; write/delete default to Ask and
surface an inline approval card (Allow / Always allow / Never) that holds the
call open with no session timeout, while read defaults to allow.

- Catalog delivered over HTTP (GET /api/v1/expert/mcp/tools), curated to
  friendly names so raw tool identifiers never reach the browser; a per-response
  hash triggers a background refetch when the catalog drifts.
- HITL state consolidated into the product-assistant store (defaults,
  per-tool preferences, pending-approval map) with SemVer version gating.
- Settings panel groups versioned tool variants into one family and points
  update hints at the newest variant's required version.
- Role inheritance is fail-closed: read-only members cannot enable or trigger
  write/delete tools and are shown why.

Use FormHeading for the section titles and ff-data-table for both the
action-type defaults and the flow-building tool list, replacing the
bespoke section/group styling and the non-standard uppercase scope
headers. Bordered table rows pair each tool with its permission control
across the row rather than leaving them to float across whitespace; tool
scope moves into a Type column.

The approval card no longer sends or renders a tool summary; the tool
name, scope and call parameters describe the action.

…n-mode

…rmissions

handleMessageResponse lived in the composer's send handler, so a reply to a
query sent from the questions card was fetched but never rendered over HTTP
(without comms-beta the reply only renders via the MQTT push handler). Fold
the response handling into handleQuery so every entry point (composer and the
question/plan cards) renders the reply without re-implementing it.

…n-mode

# Conflicts:
#	frontend/src/components/expert/components/ExpertChatInput.vue
#	frontend/src/stores/product-expert.js

…rmissions

# Conflicts:
#	frontend/src/components/expert/components/ExpertChatInput.vue

…omationsHandler integration

# Conflicts:
#	frontend/src/stores/context.js

…421)

Curate the FlowFuse platform automation tools from the handler singleton
(app.comms.platformAutomation) into the /mcp/tools catalog alongside the
flow-building tools, tagged group:'platform' so the UI routes them to their
own section with their own read/write/delete defaults. Read/write/delete
class is derived from each tool's MCP annotations; platform tools carry no
nr-assistant version window.

Replace the mid-turn approval round-trip with a stateless defer/resume flow so
the agent never stays resident waiting on a human. When a turn needs approval it
returns the approval card(s) and ends; the browser collects the decisions and
sends them back in one resume message that continues the turn.

- product-expert: track the open approval batch, resume once every card is
  answered, transport-agnostic (MQTT push or awaited HTTP reply).
- product-assistant: drop the promise-based pending-approval registry; the store
  now only records per-card outcome statuses and session grants.
- tests updated for the batch model.

…a session grant

The individual-tools table sized the tool-name column to its content, so
promoting a per-chat session grant to a saved default removed the inline
session note and its action, shrinking that column and sliding the Type
(scope) column left. Let the tool-name column absorb the free space so the
Type and Permission columns stay pinned right regardless of the note.

…lity

Send a supportsHITL flag in the Expert context. Instances predating the
human-in-the-loop tool permissions (#421) omit it, letting the agent fall back
to running flow-building tools at every scope and gating platform write/delete
tools instead of treating the user as read-only.