fix(cron): always reset session before cron runs — stateless flag inverted

[kamushadenes]

Summary

Cron jobs with stateless=true accumulate unbounded session history because the reset logic is inverted.

Current code (cmd/gateway_cron.go:92):

if !job.Stateless {
    sessionMgr.Reset(cronCtx, sessionKey)
    sessionMgr.Save(cronCtx, sessionKey)
}

This resets non-stateless jobs but skips stateless ones. The agent loop always persists messages to the session key regardless of the stateless flag, so stateless cron sessions grow indefinitely.

Observed in production: A stateless cron job accumulated 230KB of messages over ~10 runs. The model saw stale context from previous runs, causing it to skip data collection ("already done").

Fix: Remove the condition — always reset before cron execution. This is consistent with the intent of #294 (which added the reset to prevent error accumulation) and ensures stateless jobs actually start fresh.

Test plan

• Verified in production: deleting the accumulated session + running with reset fixed the behavior
• Stateless cron job starts with empty message history each run
• Non-stateless cron jobs still work (they were already being reset per bug: user_context_files becomes stale after agent_context_files update — seeding trap breaks per-user prompts #294)