ANG-012:Optional LLM enrichment (Pass B)

[yugalkaushik]

• Adds an optional second AI step ("Pass B"): once the graph has already found notes that are semantically related, it can ask Joplin's built-in AI to (1) give each note a short category label and (2) explain in plain words why two related notes are connected.
• Off by default : a new setting turns it on. It only runs on connections the graph already found, so it never scans the whole vault blindly.
• Categories and relationship explanations show up right in the graph: as a badge in note tooltips, and as the label when you hover a connection line.
• If the AI call fails or returns garbage, it retries automatically a few times, then quietly falls back to showing the graph without those labels, never blocks or breaks anything.
• Labels are cached (in memory and in the local database) so reopening the graph or making a small edit doesn't re-ask the AI for notes that haven't changed.
• Added a "Retry AI labels" setting to manually force a re-check for anything still missing a label.