Fix discount factor to avoid ambiguity with transition probabilities

[ethanelasky]

Summary

• Change discount factor from γ = 0.5 to γ = 0.8 in value iteration and policy iteration examples
• Update all calculations and resulting values to reflect the new discount factor

Motivation

The racecar MDP example uses transition probabilities of 0.5 in several places. When the discount factor is also 0.5, it creates ambiguity - students cannot easily pattern-match 0.5 and determine which represents T(s,a,s') (transition probability) and which represents γ (discount factor) in the Bellman equations.

For example, with γ = 0.5:

0.5 * [2 + 0.5 * V(s')]  # Which 0.5 is which?

With γ = 0.8:

0.5 * [2 + 0.8 * V(s')]  # Clear: 0.5 is T, 0.8 is γ

This change improves clarity for students learning MDP algorithms.

Changes

• mdp/value-iteration.md: Updated γ from 0.5 to 0.8, recalculated all values
• mdp/policies-iteration.md: Updated γ from 0.5 to 0.8, recalculated all values