Reduce AI costs: switch tagging to gpt-4o-mini, add custom base URL support

- Change AI tagging to use AI_MODELS.QUICK (gpt-4o-mini) instead of gpt-4o for
  10-15x cost reduction on classification tasks
- Add openai_base_url system setting for OpenAI-compatible providers
  (OpenRouter, Groq, Together AI, local models)
- Reset OpenAI client singleton when API key, base URL, or model changes
- Add base URL field to AI settings form with provider examples

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

src/server/services/ai-tagging.ts

@@ -16,7 +16,7 @@
  */
 
 import { prisma } from '@/lib/prisma'
-import { getOpenAI, getConfiguredModel, buildCompletionParams } from '@/lib/openai'
+import { getOpenAI, getConfiguredModel, buildCompletionParams, AI_MODELS } from '@/lib/openai'
 import { logAIUsage, extractTokenUsage } from '@/server/utils/ai-usage'
 import { classifyAIError, createParseError, logAIError } from './ai-errors'
 import {
@@ -178,7 +178,8 @@ async function getAISuggestions(
     return { suggestions: [], tokensUsed: 0 }
   }
 
-  const model = await getConfiguredModel()
+  // Use QUICK model — tag classification is simple, doesn't need expensive reasoning
+  const model = await getConfiguredModel(AI_MODELS.QUICK)
 
   // Build compact tag list for prompt
   const tagList = availableTags.map((t) => ({
@@ -294,7 +295,8 @@ async function getAISuggestionsBatch(
     return { suggestionsMap: new Map(), tokensUsed: 0 }
   }
 
-  const model = await getConfiguredModel()
+  // Use QUICK model — tag classification is simple, doesn't need expensive reasoning
+  const model = await getConfiguredModel(AI_MODELS.QUICK)
   const suggestionsMap = new Map<string, TagSuggestion[]>()
 
   // Build compact tag list (sent once for entire batch)