fix: scope read stubs by session_id on hosts without conversation_id (#1004)

[andig]

Problem

Fixes #1004. ctx_read(mode=full) could serve an [unchanged] stub on the first read of a file in a new session, pointing the model at content that session never received.

Root cause

The conversation-scope gate (core/conversation.rs) already re-delivers full content when a different conversation reads a file — but only Cursor supplied a scope id (conversation_id). Claude Code / Codex / CodeBuddy send session_id and no conversation_id, so current_conversation_id() resolved to None and the gate fell through to the legacy "allow stub" path. A new session therefore inherited the prior session's stubs.

Fix

Establish session awareness for those hosts: persist session_id into active_transcript.json and use it as the scope-id fallback (conversation_id.or(session_id)). This reuses the existing warm/cold/concurrency/subagent gate unchanged, so every tool that funnels through the shared read cache is covered.

Session semantics fall out naturally:

• new session → new id → mismatch → full re-read
• resume → same id → stub valid
• /clear → new id → re-read

Changes

• hook_handlers/observe.rs: extract session_id, persist it, and return conversation_id.or(session_id) from load_active_transcript
• core/conversation.rs: module doc updated

Tests

2 new round-trip tests; existing conversation / cache / stub-index / compaction-sync suites (26 tests) pass.

🤖 Generated with Claude Code

[yvgude]

Superseded by and integrated into #1013, including the session-scoped read-stub fix plus end-to-end hardening and full cross-platform CI. The original commit/authorship remains preserved in the integrated history. Thank you for the careful fix.