[harrison_kreps]: Fix axis

[UdohNakamura]

• Problem: np.max(..., axis=1) / np.min(..., axis=1) compares across states, not across investor types.
• Spec: Per equations (78.3) and (78.5), the max/min must be taken across types → axis=0.
• Fix: Replace axis=1 with axis=0 in the three lines below.

[HumphreyYang]

Many thanks @UdohNakamura! This looks great to me.

Hi @mmcky, this is a small (but important) fix. Should we merge it when you are ready? @UdohNakamura's change does not affect output because the two agent case is pretty special (with the same off diagnoal entries being the smallest ones). If we add a three-state two-agent exercise then it will cause issue.

[Copilot]

Pull Request Overview

This PR fixes the axis parameter in numpy max/min operations within the Harrison-Kreps model implementation. The issue was that the code was taking max/min across states (axis=1) instead of across investor types (axis=0) as required by the mathematical equations.

Key changes:

• Corrected axis parameter from 1 to 0 in three numpy operations
• Ensures proper computation according to equations (78.3) and (78.5)
• Affects both optimistic and pessimistic belief pricing functions