perf(startup): reduce storage commit amplification

[killswitch-GUI]

What changed

• defer Tantivy memory-index initialization until a memory operation needs it
• skip the redb schema write transaction when all established tables already exist
• group passive projection checkpoint advances into one adapter transaction
• preserve the existing optimistic-position and atomic rollback guarantees

Root cause

Plaintext/keyless startup still paid for durability work unrelated to journal encryption. Runtime composition eagerly created the disposable Tantivy index, redb reopened its established schema through an unconditional write transaction, and each non-applicable projection handler committed its checkpoint independently. Slow workspace-backed fsyncs amplified those commits.

Impact

On an isolated same-filesystem clone of the reported 895-event plaintext journal, time to first TUI render improved from 1.476 s to 0.618 s (58%). Journal append durability and hash-chain verification are unchanged.

Validation

• cargo test -p colossus-memory -p colossus-projection -p colossus-journal-redb -p colossus-runtime --lib
• ./scripts/check_crate_roots.sh
• cargo xtask check rust
• release-mode PTY startup benchmark on the loop-backed workspace filesystem

---

^{Need help on this PR? Tag @codesmith with what you need. Autofix is enabled.}

[chatgpt-codex-connector]

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: 06178f21c5

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

• Open a pull request for review
• Mark a draft as ready
• Comment "@codex review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".