fix: use CHAT_SETTING_LIMITS for max_tokens instead of hardcoded values

[KushalLukhi]

Problem

The OpenAI and Azure chat routes had hardcoded max_tokens logic with TODO comments indicating this needed to be fixed. Only gpt-4-vision-preview and gpt-4o were explicitly set to 4096 tokens, while other models had no limit set (null).

Solution

Updated both routes to use CHAT_SETTING_LIMITS to look up the appropriate MAX_TOKEN_OUTPUT_LENGTH for each model, ensuring consistent token limits across all models.

Changes

• app/api/chat/openai/route.ts: Import CHAT_SETTING_LIMITS and use it for max_tokens
• app/api/chat/azure/route.ts: Import CHAT_SETTING_LIMITS and use it for max_tokens
• Removed TODO comments that were tracking this issue

Benefits

• All models now use their configured MAX_TOKEN_OUTPUT_LENGTH from CHAT_SETTING_LIMITS
• No more hardcoded model-specific logic
• Future model additions automatically use correct token limits
• Consistent behavior between OpenAI and Azure routes

Testing

• TypeScript compilation passes
• All existing models continue to use their configured token limits
• gpt-4-vision-preview and gpt-4o still use 4096 tokens as before
• Other models (gpt-4-turbo-preview, gpt-4, gpt-3.5-turbo) now correctly use their limits

Fixes technical debt noted in TODO comments.