fix(app): queue concurrent approval requests instead of overwriting them

[addyCooks]

Closes #1156

Description

The three "ask the user" slots question-queue, tool-approval-queue and
tool-confirm-queue each stored a single resolver in a useRef. A second
caller arriving before the first was answered overwrote it, and that first
promise could then never settle.

tool-executor starts up to MAX_CONCURRENT_AGENTS (5) subagents in one turn
and awaits them with Promise.allSettled, so one stranded caller meant the batch
never resolved and the turn never ended. Escape did not help either: the subagent
was parked in an await that is not abort-aware, so recovery meant killing the
process.

Each slot now backs onto a FIFO shared by all three. Requests queue in arrival
order, the head is presented, and answering it resolves that caller and promotes
the next so every caller settles exactly once with its own answer.

The three blocks were near-identical, so they collapse into one useHandlerQueue
helper used three times; the change is a net reduction in the file. The hook's
public shape is unchanged, so App.tsx, chat-input.tsx and interactive-app.tsx
needed no edits.

Scope

Two items from the issue's suggested fix are deliberately not here:

• A "1 of N" indicator. It changes the hook's public shape and pulls in
  chat-input.tsx and interactive-app.tsx. Worth doing, but it is UX on top of
  the fix rather than part of it.
• An abort-aware signal(). That changes createGlobalHandlerSlot's
  signature and every call site, including subagent-executor.ts:739 where the
  sub-agent tool bridge work for ACP: thread sub-agent tool results back so every call is announced and settles with its real status #1105 is also editing. Queuing alone removes the
  hang; abort handling is a clean follow-up once that lands.

Type of Change

• Bug fix
• New feature
• Breaking change
• Documentation update

Changeset

• Added a changeset (pnpm changeset) describing this change for the changelog

Testing

Automated Tests

• New features include passing tests in .spec.ts/tsx files
• All existing tests pass (pnpm test:all completes successfully)
• Tests cover both success and error scenarios

Manual Testing

• Tested with Ollama
• Tested with OpenRouter
• Tested with OpenAI-compatible API
• Tested MCP integration (if applicable)

Checklist

• If this was for an open issue, I was assigned to it
• Code follows project style guidelines
• Self-review completed
• Documentation updated (if needed)
• No breaking changes (or clearly documented)
• Appropriate logging added using structured logging (see CONTRIBUTING.md)