cleanup(fflag): delete dead enable_causal_intervention chain — [FFLAG-011]

Write-only flag chain: DqnTrainerConfig.enable_causal_intervention →
local causal_enabled → CausalInterventionConfig.enabled. Zero readers
anywhere — causal intervention buffers are unconditionally allocated
per the existing "ALWAYS allocated (one production path)" comment.
Deleted all three sites + 2 construction setters. Weight/interval/
market_dim fields retained (those are read).

crates/ml/src/cuda_pipeline/gpu_dqn_trainer.rs

@@ -300,8 +300,6 @@ pub struct GpuDqnTrainConfig {
     pub ensemble_disagreement_penalty: f32,
     /// #21 Stochastic depth drop probability. 0.0 = disabled, 0.2 = 20% drop per layer.
     pub stochastic_depth_prob: f32,
-    /// #34 Enable causal intervention training.
-    pub enable_causal_intervention: bool,
     /// #34 Causal regularization weight.
     pub causal_weight: f64,
     /// #34 Run causal intervention every N training steps.
@@ -365,7 +363,6 @@ impl Default for GpuDqnTrainConfig {
             stochastic_depth_prob: 0.0,
             pruning_epoch: 50,
             pruning_fraction: 0.7,
-            enable_causal_intervention: true,   // always on
             causal_weight: 0.1,
             causal_intervention_interval: 50,
             enable_gradient_vaccine: true,     // always on
@@ -429,7 +426,6 @@ pub struct FusedTrainScalars {
 /// #34 Causal intervention configuration.
 #[derive(Clone)]
 pub(crate) struct CausalInterventionConfig {
-    pub enabled: bool,
     pub weight: f32,
     pub interval: usize,
     pub market_dim: usize,
@@ -437,7 +433,7 @@ pub(crate) struct CausalInterventionConfig {
 
 impl Default for CausalInterventionConfig {
     fn default() -> Self {
-        Self { enabled: false, weight: 0.0, interval: 10, market_dim: 42 } // market_dim overridden at construction
+        Self { weight: 0.0, interval: 10, market_dim: 42 } // market_dim overridden at construction
     }
 }
 
@@ -6361,8 +6357,7 @@ impl GpuDqnTrainer {
             }
         };
 
-        // #34 Causal intervention buffers
-        let causal_enabled = config.enable_causal_intervention;
+        // #34 Causal intervention buffers (always enabled — one production path)
         let causal_market_dim = config.market_dim;
         let causal_wt = config.causal_weight as f32;
         let causal_intv = config.causal_intervention_interval;
@@ -7467,7 +7462,6 @@ impl GpuDqnTrainer {
             causal_mean_reduce_kernel: causal_mean_reduce_kernel_fn,
             causal_mean_scratch,
             causal_config: CausalInterventionConfig {
-                enabled: causal_enabled,
                 weight: causal_wt,
                 interval: causal_intv,
                 market_dim: causal_market_dim,

crates/ml/src/trainers/dqn/fused_training.rs

@@ -403,7 +403,6 @@ impl FusedTrainingCtx {
             stochastic_depth_prob: hyperparams.stochastic_depth_prob as f32,
             pruning_epoch: hyperparams.pruning_epoch,
             pruning_fraction: hyperparams.pruning_fraction as f32,
-            enable_causal_intervention: true,
             causal_weight: hyperparams.causal_weight,
             causal_intervention_interval: hyperparams.causal_intervention_interval,
             enable_gradient_vaccine: true,