Refactor: Database functions and Course querying functions

The choice of database when creating the rework was MongoDB, as it supported free usage limits well above what is required for UIGrades data; however, course queries go to the database on every filter change and search change (on the `/courses` route at uigrades.vercel.app/vourses). Research whether pulling all courses, memoizing them, and running filtering functions on the repository side is quicker (or not) than making requests to MongoDB on every query change. This issue isn't very high priority since there aren't that many items in the database, so queries to MongoDB are relatively quick.

Identify where the `/courses` (shown as `/vourses`) page triggers filter/search updates and where each event calls MongoDB. Trace those query functions and measure current request frequency/latency. Then compare with a path that memoizes loaded courses and applies repository-side filtering, using any existing benchmarking or logs. Done means a clear recommendation or proof-of-concept plan with measured results before refactoring.

Do mô hình lập chỉ mục viết ra từ nội dung của issue.