Feature clipping is a simple feature-engineering trick to reduce the impact of extreme values by capping them at chosen thresholds. In this task, you’ll clip each numeric feature column using per-feature lower and upper bounds.

Requirements

Implement the function

Rules:

• Clip each value with: $x'_{i,j} = \min(\max(x_{i,j}, \text{lower}_j), \text{upper}_j)$
• Do not modify the input X in place; return a new clipped 2D array.
• Use only NumPy and Python built-in libraries (no pandas).
• Treat each column independently using its corresponding bounds.
• Keep the output values as floating point numbers.

Example

Output: