Fix GPT-5 models by removing max_tokens parameter

updated the code for llm abs class

Fix OpenAI default model to use gpt-5-mini-2025-08-07

Add provider-model validation to prevent API mismatch errors

Respect user's preferred LLM provider and model across all endpoints

Add user-scoped sequential numbering to resumes, jobs, and matches pages

Aggressively target 65-75% keyword match for ATS score improvement

Change IDs to user-scoped sequential numbers

Fix LLM returning JSON metadata instead of resume text

Revert "Implement smart OpenAI model routing for 40-50% cost reduction"

Fix JSON parsing errors with robust error handling and fallback

Implement smart OpenAI model routing for 40-50% cost reduction

Sort matches by date and add auto-refresh after resume operations

Fix cover letter voice to use consistent first person

Fix onboarding tour, OAuth errors, and UI refresh issues