perf(shared-fs): remove unused recursive history depths

[peerbit-org]

Summary

Small master-based cleanup, independent of held #331/upstream upgrade: remove unused recursive history-depth computation and private depth maps. Keep head filtering and all stored-causalDepth/non-delete/ID winner rules unchanged. No new cache, API/schema change, dependency change or replication-policy change.

The new pre-change regression exposed Maximum call stack size exceeded for a 20,000-event reverse-ordered naming history. Removing the unused recursion fixes it; both naming and content deep-chain cases now pass.

Verification

• 23 new deterministic cases pass: exact heads under permutations, forks/merges, missing/retired ancestors, stored-depth and ID ordering, naming non-delete bias, existing malformed-cycle behavior, deep reverse chains.
• Independent code/test review, library build, ESLint and formatting pass.
• Package dry-run: 2,203,473 unpacked bytes / 492,964 compressed, under budget, no test files included. No publish.
• A matched synthetic warm-head microbenchmark (60 samples after10 warmups per case; separate processes, baseline first) showed 10k-row naming p50 1.510→0.695ms, content1.711→0.802ms. Component-only, no IO/network/mount claim, no guaranteed speedup; O(history) remains. Raw evidence is preserved locally.

HOLD — strict full run failed

The full library suite used the opt-in runner/reporter from #340 via explicit config path, without changing its budgets. It reported542 eventual passes /10 skipped, but correctly exited1 on one actual persistent-multi-writer retry (missingInstrumentation:0). Do not treat that as a clean run or merge this draft.

First failure: expected undefined to be defined from persistedReceiptPeerSession(remoteHash) at persistent-multi-writer.test.ts:100, reached from line400's Promise.all index1 after restart/convergence and before prepareForDisposal. The remote identity was not captured. This resembles the prior older-cohort readiness failure category; it is not a new proof of latest-cohort failure, data loss, or a proven identical root cause.

All23 new tests and existing shared-fs31 /GC14 /compaction6 /durable-disposal22 cases passed without retries. No local or CI rerun, timeout inflation, assertion weakening or check bypass. Ordinary CI on this branch does not yet include #340's strict gate; a green result there does not clear this recorded hold.

Local raw evidence: /private/tmp/peerbit-performance-20260905/unused-depth-library-strict.log (first failure line230), unused-depth-baseline-regressions.log, unused-depth-after-regressions.log, unused-depth-baseline.raw.log, unused-depth-after.raw.log.

All git/GitHub actions as verified peerbit-org. No upstream writes or release.