refactor(sdk): share aggregator shape across hotspots roll-ups

[willwashburn]

Closes #340.

Summary

Three small clean-ups in crates/relayburn-sdk/src/analyze/hotspots.rs:

• Shared aggregator shape. aggregate_by_file, aggregate_by_bash, and aggregate_by_subagent had near-parallel "filter → accumulate by key → sort by cost desc" boilerplate. Extract a private generic aggregate(…) helper parameterized by key/init/accumulate/cost closures; each public function now declares only the bits that actually differ. aggregate_by_bash_verb keeps its own loop because it tracks distinct hashes and per-verb examples on top of the basic shape.
• f64::total_cmp for monotonic cost ordering. Replace every partial_cmp(...).unwrap_or(Ordering::Equal) cost comparator in this file with total_cmp — drop-in for our positive-cost values and removes the silent-Equal fallback for NaN. Covers the file/bash/subagent aggregates and both sorts in the bash-verb path.
• One pass over session_turns in attribute_hotspots. attribute_session already iterates the session's turns; have it accumulate the per-session grand total in that same pass via cost_for_turn and return it on SessionAttribution. The caller drops the second for t in &session_turns { … } loop. Behavior is unchanged — totals still route through the canonical cost_for_turn so source-specific reasoning billing (Codex included_in_output, separate reasoning tariffs, etc.) stays in lock-step with cost.rs.

No public-API changes. Existing analyze::hotspots tests cover deterministic sort, Codex reasoning, and the grand_total + unattributed = session_grand invariant.

Test plan

• cargo test -p relayburn-sdk --lib analyze::hotspots (14 passed)
• cargo test --workspace (775 passed, 0 failed)
• cargo clippy -p relayburn-sdk --all-targets — no new warnings on the touched code

https://claude.ai/code/session_01Wmh6iHN5aYaPKhFdR9bA14

---

Generated by Claude Code

[coderabbitai]

No actionable comments were generated in the recent review. 🎉

ℹ️ Recent review info

⚙️ Run configuration

Configuration used: Organization UI

Review profile: CHILL

Plan: Pro Plus

Run ID: 954b4e4a-940a-4f00-967b-39dfbf91b72a

📥 Commits

Reviewing files that changed from the base of the PR and between 7da0599 and b489ef1.

📒 Files selected for processing (1)

• crates/relayburn-sdk/src/analyze/hotspots.rs

---

📝 Walkthrough

Walkthrough

This PR consolidates hotspot analysis logic: session grand totals are computed once in the attribution loop and reused downstream, three parallel aggregation functions are unified under a generic helper, and float comparisons are switched to deterministic ordering for consistent output.

Changes

Hotspot Attribution and Aggregation Refactor

Layer / File(s)	Summary
Session Attribution Data Shape crates/relayburn-sdk/src/analyze/hotspots.rs	SessionAttribution struct gains a new grand_total: f64 field to hold session-level cost sums.
Grand Total Computation in Attribution crates/relayburn-sdk/src/analyze/hotspots.rs	attribute_session accumulates each turn's cost_for_turn(...).total into a grand_total variable during the per-turn loop; empty-turns path initializes grand_total to 0.0.
Grand Total Reuse in Hotspots crates/relayburn-sdk/src/analyze/hotspots.rs	attribute_hotspots derives each session's grand_cost from session_result.grand_total instead of recalculating it in a separate pass.
Generic Aggregation Helper crates/relayburn-sdk/src/analyze/hotspots.rs	New internal generic aggregate helper iterates attributions, groups rows by extracted key, accumulates shared metrics (total_cost, total_count), and sorts by total_cost descending.
Refactored Aggregation Functions crates/relayburn-sdk/src/analyze/hotspots.rs	aggregate_by_file, aggregate_by_bash, and aggregate_by_subagent are rewritten to use the aggregate helper, preserving file-target filtering, args-hash collapsing, and (unknown) fallback behavior.
Deterministic Float Comparisons crates/relayburn-sdk/src/analyze/hotspots.rs	Bash-verb example and final row sorting switched total_cost comparators from partial_cmp(...).unwrap_or(...) to f64::total_cmp.

Estimated code review effort

🎯 3 (Moderate) | ⏱️ ~25 minutes

Possibly related PRs

• AgentWorkforce/burn#386: Both PRs refactor hotspot aggregation/attribution inside relayburn-sdk analyze code; #386 modifies summary/hotspots aggregation while this PR refactors the underlying hotspots.rs logic.
• AgentWorkforce/burn#291: Both PRs modify hotspots attribution and aggregation functions (attribute_hotspots, aggregate_by_file/bash/subagent, session grand-total handling).

Poem

🐰 Three loops become one, grand totals aligned,
Aggregation patterns in harmony bind,
Float comparisons settled with total_cmp grace,
Hotspot refactors now run a faster race! ✨

🚥 Pre-merge checks | ✅ 5

✅ Passed checks (5 passed)

Check name	Status	Explanation
Title check	✅ Passed	The title clearly summarizes the main refactoring change: introducing a shared generic aggregator shape across hotspot rollup functions, which is the primary objective of the PR.
Description check	✅ Passed	The description is directly related to the changeset, detailing the three clean-ups made to hotspots.rs and referencing the closed issue #340, providing context for the refactoring.
Linked Issues check	✅ Passed	All primary objectives from issue #340 are addressed: generic aggregator shape introduced [#340], f64::total_cmp used for cost ordering [#340], and session_turns loop consolidated [#340].
Out of Scope Changes check	✅ Passed	All changes align with issue #340 scope: generic aggregate helper extraction, f64::total_cmp migration, and session loop consolidation are the three stated objectives.
Docstring Coverage	✅ Passed	Docstring coverage is 100.00% which is sufficient. The required threshold is 80.00%.

_{✏️ Tip: You can configure your own custom pre-merge checks in the settings.}

✨ Finishing Touches

📝 Generate docstrings

• Create stacked PR
• Commit on current branch

🧪 Generate unit tests (beta)

• Create PR with unit tests
• Commit unit tests in branch claude/resolve-github-issue-0jdvd

---

_{Comment @coderabbitai help to get the list of available commands and usage tips.}

[devin-ai-integration]

✅ Devin Review: No Issues Found

Devin Review analyzed this PR and found no potential bugs to report.

View in Devin Review to see 3 additional findings.