Bug: nodes.hostname stores display names, not actual hostnames

[mrveiss]

Problem

The nodes.hostname column in the PostgreSQL nodes table stores human-readable display names (e.g., 00-SLM-Manager, 01-Backend) rather than actual OS hostnames (autobot-slm, MV-Stealth) or Ansible inventory hostnames (autobot-slm, autobot-host).

This caused the entire System Updates feature to silently fail (#1789) because code assumed hostname was an actual hostname usable for Ansible --limit and host resolution.

Current state

node_id	DB hostname (display)	OS hostname	Ansible inventory name
00-SLM-Manager	00-SLM-Manager	autobot-slm	autobot-slm
01-Backend	01-Backend	MV-Stealth	autobot-host
02-Frontend	02-Frontend	01-Backend	autobot-frontend
npu-worker	npu-worker	npu-worker	autobot-npu
04-Databases	03-Databases	03-Databases	autobot-database
03-AI-Stack	04-AI-Stack	04-AI-Stack	autobot-aiml
browser-automation	05-browser-automation	05-browser-automation	autobot-browser

Impact

Any code that uses Node.hostname expecting an actual hostname will break. The workaround in #1789 (inventory parsing + IP fallback) is fragile. Other subsystems (deployments, health checks, etc.) that build Ansible commands from Node.hostname may have the same latent bug.

Proposed fix

Add ansible_name column to nodes table, populated from the inventory during enrollment. Keep hostname as a display name but rename to display_name for clarity. Audit all callers of Node.hostname.

Discovered During

Working on #1789

[mrveiss]

Merged in PR #1932. Added ansible_name column + Node.ansible_target property. All Ansible --limit callers updated.