[FEATURE]:Session compaction includes already-summarized messages, wasting tokens

[Liuhanlin23]

Feature hasn't been suggested before.

• I have verified this feature I'm about to request hasn't been suggested before.

Describe the enhancement you want to request

Description:
The session compaction logic in compaction.ts currently includes all historical messages every time it performs compaction, even when previous compaction summaries already exist. This causes unnecessary processing and token waste.
I noticed that the prune() method already implements this optimization, recognizing that messages before a summary point don't need reprocessing.
If you agree this optimization is needed, I would like to be assigned to handle and fix this issue.

[github-actions]

This issue might be a duplicate of existing issues. Please check:

• Bug: Watchdog reminder causes excessive token amplification (O(N) growth) #11142: Bug about watchdog reminder causing excessive token amplification (O(N) growth) - similar reprocessing of already-included messages causing token waste
• [FEATURE] Sliding window context management for long-running sessions #4659: [FEATURE] Sliding window context management for long-running sessions - proposes inception messages to avoid reprocessing historical context
• [FEATURE] Recursive Language Model (RLM) Context Management - Context as External Environment #11829: [FEATURE] Recursive Language Model (RLM) Context Management - suggests model should query context programmatically rather than include all messages to eliminate redundant reprocessing
• [FEATURE]: Allow pinning important instructions/messages per session to keep them after compaction. #8932: [FEATURE] Allow pinning important instructions/messages per session to keep them after compaction - addresses context loss and provides mechanism to preserve critical context
• [FEATURE]: Recursive Language Model as alternative to Compaction #8455: [FEATURE] Recursive Language Model as alternative to Compaction - proposes external context storage model

These issues all converge on the same root cause: compaction logic reprocessing messages that have already been summarized, causing token waste and context degradation.

Feel free to ignore if none of these address your specific case.