Harden hosted siftd runtime and deployment baseline

[t3chn]

Summary

• add configurable hosted retention window to siftd with a 30-day default
• apply retention in the shared sync runtime and remove orphaned events in both Postgres and SQLite stores
• add regression coverage for retention and standalone store cleanup behavior
• add single-node deployment assets for siftd: systemd unit, environment template, and operator runbook

Issue ID

• N/A

Test Plan

• go test ./...
• jq . project.manifest.json >/dev/null
• git diff --check

[chatgpt-codex-connector]

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: 23eae9f8e3

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".

[chatgpt-codex-connector]

Make retention and event rebuild atomic

Applying retention in a separate committed transaction before rebuilding events can leave canonical state inconsistent when any later step fails (for example ListArticlesForClustering/ReplaceEvents returning an error). In that case, old articles/event_articles are already deleted but events is still the previous snapshot, so hosted reads can continue serving out-of-retention events and event payloads whose supporting article rows no longer exist until a later successful sync. This regression was introduced by the new store.ApplyRetention call path and is likely to violate the intended retention/provenance guarantees under transient DB failures.

Useful? React with 👍 / 👎.