fix(tui): bound the reasoning stream to a tail and skip it when collapsed

[addyCooks]

Closes #1329

Description

StreamingReasoning called wrapWithTrimmedContinuations(reasoning.trimEnd(), textWidth) on the entire accumulated reasoning string on every token flush, with no bound and no early exit. StreamingMessage already solved this with computeStreamingTail; the reasoning component never adopted it. Worse, reasoningExpanded defaults to false, so the full re-wrap plus calculateTokens ran on every flush for a panel that renders nothing.

Fix:

• Reuse computeStreamingTail (already exported from streaming-message.tsx) so each flush is O(tail) rather than O(history).
• Skip the wrap and the token count entirely when expand === false.

No new helper and no new file the bounded-tail logic stays in one place.

Measured, 30 flushes over a ~150KB trace (same shape as the harness in the issue):

	before	after
collapsed	2757ms total 91.9ms/flush	348ms 11.6ms/flush
expanded	2977ms total 99.2ms/flush	757ms 25.2ms/flush

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

[github-actions]

nc-review: nothing to raise

@addyCooks — nothing to raise from the automated review.

The PR ports the bounded-tail wrap pattern from StreamingMessage into StreamingReasoning and short-circuits the wrap and token count when expand === false. The reuse is direct — same helper, same MAX_LINES, same sliced || lines.length > MAX_LINES truncated-or-not logic — so the symmetry is preserved and no new file is added. The three new tests target the actual behaviours that regressed under the old code (tail across flushes, collapsed-default rendering nothing) and the changeset is present with the right workspace package name. Ready to merge.

---

_{🔴 blocking · 🟠 a reviewer would ask for a change · ⚪ optional}

_{Automated code review — correctness, security, design, tests, plus duplicates and scope. A human still decides; this is not a substitute for review and is not exhaustive. The required status checks separately cover lint, formatting, types, unused dependencies, the test suite and the build. This bot never merges. Maintainers can rerun with /re-review.}

[will-lamerton]

Great work @addyCooks :)

[addyCooks]

Thanks @will-lamerton : )