fix: eliminate 3x SQLite message duplication in gateway sessions

[teknium1]

Summary

Fixes #860 — SQLite session transcripts accumulated duplicate messages (3-4x token inflation).

Root Cause

Three separate code paths all wrote to the same state.db with no deduplication:

1. _log_msg_to_db() — wrote each message individually right after messages.append()
2. _flush_messages_to_session_db() — re-wrote ALL new messages at every _persist_session() call (~18 exit points), with no tracking of what was already written
3. Gateway append_to_transcript() — wrote everything a third time after the agent returned

Since load_transcript() prefers SQLite over JSONL, the inflated data was loaded on every session resume, causing proportional token waste.

Fix

run_agent.py:

• Remove _log_msg_to_db() method and all 16 call sites (redundant with the flush mechanism)
• Add _last_flushed_db_idx tracking in _flush_messages_to_session_db() so repeated _persist_session() calls only write truly new messages
• Reset flush cursor on compression (new session ID)

gateway/session.py:

• Add skip_db parameter to SessionStore.append_to_transcript() — when True, writes JSONL only

gateway/run.py:

• Pass skip_db=True when the agent already persisted messages to SQLite
• JSONL backup writes still happen (backward compat)

Verification

Live-tested with hermes chat: a 12-message session with 10 tool calls produces exactly 12 SQLite rows with zero duplicates (previously would have been 36-48).

Tests

• 9 new tests in tests/test_860_dedup.py covering:
  • Flush deduplication (repeated calls write nothing new)
  • Incremental flush (only new messages written)
  • Multiple _persist_session calls (no duplication)
  • Compression reset (flush cursor resets for new session)
  • skip_db=True prevents SQLite writes
  • skip_db=False (default) writes to both stores
  • _last_flushed_db_idx initialization
• Updated test_interrupt.py to remove reference to deleted _log_msg_to_db
• Full suite: 2869 passed, 0 failed