fix: address review feedback on profiler flamegraph/icicle

[Oaklight]

Summary

Addresses all actionable and non-blocking review feedback from #166:

• Raise ValueError when sort_by/limit passed with flamegraph/icicle styles (flagged by all 3 reviewers)
• Remove dead zoomStack code — reset renders directly from root data (flagged by all 3 reviewers)
• Unify total time source — _build_flame_html now uses total_tt consistently with _extract_call_tree (milo)
• Add edge tuple format comment for CPython version differences (clementine)
• Clarify tooltip label — "Cumulative (path)" instead of ambiguous "Cumulative" (milo, elena)
• Add empty-profile test for flamegraph + sort_by/limit rejection tests (clementine)

Test plan

• make test-profiler — 68 tests pass (3 new)
• pre-commit run --all-files — all hooks pass
• CI lint + test

[clementine-oaklight]

Review — PR #167: fix: address review feedback on profiler flamegraph/icicle

CI: ✅ Green (lint + 3.10–3.13)
Verdict: LGTM — clean follow-up, all feedback items addressed.

---

What's fixed

#166 feedback	Resolution
sort_by/limit silently ignored for flamegraph/icicle	ValueError raised when either is passed with non-table style
zoomStack dead code	Removed entirely — click-to-zoom re-renders directly, reset re-renders from root
total_tt vs self.total_time dual access	_build_flame_html now reads stats.total_tt directly, matching _extract_call_tree
Edge tuple format undocumented	Comment added: # edge is (cc,nc,tt,ct) or (nc,ct) depending on CPython version
No empty-profile test	test_flamegraph_empty_profile added
Tooltip ambiguity on cumtime meaning	Label changed to "Cumulative (path):"

Three new tests cover the empty-profile path and both ValueError cases. +26/−6 across 2 files — focused and minimal. Nothing to flag.

[elena-oaklight]

CI green across lint + 3.10–3.13. Clean, focused follow-up that addresses all actionable feedback from #166 — nothing missed.

• sort_by/limit now raise ValueError for flamegraph/icicle — guard is placed before _extract_call_tree(), so no wasted work. Tests cover both params independently.
• zoomStack removed — all three references gone (declaration, push in onFrameClick, clear in reset). onFrameClick simplified to just render([node], node.cumtime). Clean.
• total_tt unified — _build_flame_html now uses cast(Any, self._stats).total_tt instead of self.total_time, consistent with _extract_call_tree. Fixes the CPU-vs-wall-clock divergence milo flagged.
• Edge tuple comment — documents the (cc,nc,tt,ct) vs (nc,ct) CPython version difference inline.
• Tooltip label — "Cumulative (path)" clarifies the edge-specific semantics.
• Empty-profile test — good edge case addition.

+26/−6, no issues. LGTM 👍

[milo-oaklight]

All six items from #166 addressed cleanly:

1. sort_by/limit now raise ValueError for flamegraph/icicle — good, explicit beats silent.
2. zoomStack removed — dead var, push in onFrameClick, and reset-handler clear all gone. Bonus: dropped the unused e parameter too.
3. total_time → total_tt in _build_flame_html — CPU time now consistent with _extract_call_tree. Comment makes the intent clear.
4. Edge tuple format comment added at the extraction site.
5. Tooltip label → "Cumulative (path)" — disambiguates from function-global tottime.
6. Three new tests — empty profile, sort_by rejection, limit rejection. All three assert the right error message.

CI green across lint + 3.10–3.13. Nothing else to flag — straightforward follow-up.