Chore/simple nn test

[TeddyCr]

Summary

• Add simple test case for MLP to improve reasoning about the implementation

• Fix single neuron output handling in MLP implementation

• Expand .gitignore with standard Python patterns from GitHub repo.

  Changes

  Testing improvements:

  • Added test/test_nn.py with a simple MLP test that validates both forward and backward passes with known expected values
  • Test uses a 1 hidden layer network with a single neuron (1 → 1 → 1) with manually set weights and biases to verify gradient computation
    [This helped me resonate a bit better about the code implementation -- thought it could help others.]

  Bug fix:

  • Fixed output handling for single-neuron MLPs by moving the list unwrapping logic from Layer.__call__() to MLP.__call__()
    [I believe layer(x) should always return a list[Value] based on the expected inout in Neuron activation computation (zip(w, v)).]

  Housekeeping:

  • Updated .gitignore using https://github.com/github/gitignore/blob/main/Python.gitignore to exclude Python build artifacts, virtual environments, IDE files, and distribution packages