Try probabilistic mode of disabling genes

[MarkZH]

Instead of a binary on/off switch, each gene has a probability of being used for any given turn.

1. Create a double member field probability of use that is clamped to the range [0, 1].
2. At the beginning of a turn, generate a random number threshold in the range [0, 1].
3. Instead of checking active() for gene methods, check that probability >= threshold before running the method.

This will require rescaling the pawn score for thinking output each move (and quickly). Maybe save the score for each gene after initializing and rescaling for disabled genes on each turn.

[MarkZH]

The probability field should be an Interpolated_Gene_Value.

[MarkZH]

Different idea: two components (activation_start and activation_end) that define an interval in the range [0, 1] during which the gene is active.

bool active(double game_progress) const noexcept
{
    return activation_start <= game_progress && game_progress <= activation_end;
}

Mutation restriction: -0.5 <= activation_start <= activation_end <= 1.05 (0.05 margins to allow for being on since the first move and on until the last move. If a mutation puts start and end in wrong order, change the non-mutated value to accommodate the restrictions.