gradient_update: gradient_update
In KennethEnevoldsen/SiRToM: SirToM: Simulation in R: Theory of Mind.

k-ToM calculates the gradient between parameter estimates and choice probability estimates

1 2	gradient_update(opponent_prev_hidden_states, params, mean, param_mean, reverse_choices, opponent_level, opponent_player, p_matrix.)

`opponent_prev_hidden_states`	opponents previous hidden states
`mean`	the mean of k-ToM's choice probability estimate of its opponent
`param_mean`	the means of k-ToM's parameter estimates
`reverse_choices`	a vector of k-ToM's 1 opponent's choice and 2 own choice. Inserted as choices when simulating the opponent
`opponent_level`	the level of the opponent for which the gradient is calculated
`opponent_player`	the reverse player role of k-ToM's own. Inserted as player role when simulaitng the opponent
`p_matrix`	a given 2-by-2 payoff matrix