Chore: add CI check for Ansible slm_agent role code drift #1629

Problem

The SLM agent code exists in two locations:

• autobot-slm-backend/slm/agent/ — canonical source (actively developed)
• autobot-slm-backend/ansible/roles/slm_agent/files/slm/agent/ — Ansible role copy (deployed to new nodes)

These copies WILL drift again after #1623 synced them. Any future change to the canonical agent will silently leave the Ansible copy stale until someone manually notices.

Impact

• Severity: Medium — Newly provisioned nodes get outdated agent code
• This has already caused issues: Bug: Ansible slm_agent role has stale agent code — missing crash-loop detection #1623 found the Ansible copy was missing crash-loop detection (Bug: no alerting for systemd service crash loops on fleet nodes #1604), error context (feat(slm): per-service health monitoring in SLM dashboard #1019), and refactored helpers ([Code Quality] Fix 600 Long Functions (Exceeding Recommended Length) #620)

Proposed Fix

Option A (recommended): CI check
Add a GitHub Actions step or pre-commit hook that compares the two directories:

diff -r autobot-slm-backend/slm/agent/ \
       autobot-slm-backend/ansible/roles/slm_agent/files/slm/agent/ \
       --exclude='__pycache__' --exclude='*_test.py'

Fail CI if they diverge.

Option B: Eliminate the copy
Modify deploy-slm-agent.yml to rsync from the canonical source instead of bundling a static copy in the role.

Option C: Symlink
Replace the Ansible role copy with a symlink to the canonical source (requires adjusting the Ansible role to resolve symlinks).

Discovered During

Implementing #1623 (sync stale agent code).