feat(compaction): add dormant hard compaction planner

[Hmbown]

Summary

• add a default-disabled HardCompactionConfig and pure plan_hard_compaction helper
• derive the hard-compaction plan from the existing plan_compaction machinery so recent tail, working-set pins, user-text fallback, and tool-call/result pair guarantees stay intact
• add focused tests for too-short histories, recent-tail preservation, and the tool-pair boundary risk called out in review

Stewardship

Refs #2522. This harvests the safe planning layer from @HUQIANTAO's hard-compaction proposal without enabling hard compaction, adding user config, or adding a message-rewrite/LLM execution path. The broader runtime behavior remains deferred until tool-pair atomicity, anchors, working-set preservation, and cache/prefix metadata semantics are fully specified and tested.

Verification

• cargo test -p codewhale-tui compaction --locked
• cargo fmt --all
• ./scripts/release/check-versions.sh
• ./scripts/release/check-ohos-deps.sh
• cargo clippy -p codewhale-tui --bin codewhale-tui --locked -- -D warnings
• cargo fmt --all -- --check
• git diff --check
• cargo check --workspace --all-features --locked

[greptile-apps]

Hmbown has reached the 50-review limit for trial accounts. To continue receiving code reviews, upgrade your plan.

[gemini-code-assist]

Code Review

This pull request introduces a default-disabled hard-compaction planner that identifies the summarizable middle of a long conversation while preserving the recent tail, tool-call/result pairs, and working-set pinning. Feedback suggests optimizing the generation of preserved_indices in plan_hard_compaction by directly collecting soft_plan.pinned_indices instead of manually computing the complement of summarize_indices, which reduces complexity and avoids unnecessary allocations.

Important

The consumer version of Gemini Code Assist on GitHub is being sunset. Starting June 18, 2026, new organization installations will be blocked, and all code review activity will officially cease on July 17, 2026.
For more details on the timeline and next steps, please review the Help Documentation.

[gemini-code-assist]

The preserved_indices are the exact complement of summarize_indices in the compaction plan, which is already computed and stored in soft_plan.pinned_indices. Instead of collecting summarize_indices into a temporary BTreeSet and performing a linear scan with .contains() checks, you can directly consume and collect soft_plan.pinned_indices into a Vec. This avoids unnecessary allocations and reduces the complexity from O(N log N) to O(N).

Suggested change

	let summarized: BTreeSet<_> = soft_plan.summarize_indices.iter().copied().collect();
	let preserved_indices = (0..messages.len())
	.filter(|idx| !summarized.contains(idx))
	.collect();
	let preserved_indices = soft_plan.pinned_indices.into_iter().collect();