Implement an in-memory SQL-like table

Implement a simplified in-memory SQL-like database for a single table. You should support the following operations on a table with a fixed schema (a list of column names). Each row is a mapping from column name to value (you may assume values are strings or integers and can be compared). Requirements: 1. **INSERT**: Insert a row providing a value for each column. 2. **SELECT with projection**: Return rows while selecting only a specified subset of column names (e.g., `SELECT colA, colC`). 3. **Filtering**: Support a `WHERE` clause that filters rows by a condition on a column (at minimum, equality like `colX = value`; if you choose, you may support additional comparisons). 4. **Ordering**: Support `ORDER BY <column>` (ascending; optionally support descending). Clarify/handle: - Invalid column names in SELECT/WHERE/ORDER BY. - Rows with missing columns (if disallowed, validate on insert). - Output format: a list of projected rows in the final order. Design an API (e.g., `insert(row)`, `select(columns, where_clause, order_by)` or parse SQL strings if preferred) and implement the functionality efficiently for typical interview constraints (up to tens or hundreds of thousands of rows).