cleanup(fflag): delete dead use_noisy_nets + use_distributional — [FFLAG-013a]

GpuExperienceCollectorConfig.use_noisy_nets and use_distributional were
both declared with "always enabled" comments on their default=true
setters. Zero conditional readers anywhere — the GPU kernel always
runs NoisyNet exploration and C51 distributional RL. Deleted fields +
defaults + training_loop setters. Kept noisy_sigma_init, num_atoms,
v_min, v_max since those are read.

enable_action_masking left for human review — has real gating chain
across collector config + trainer struct + training loop.

crates/ml/src/cuda_pipeline/gpu_experience_collector.rs

@@ -200,12 +200,8 @@ pub struct ExperienceCollectorConfig {
     pub q_clip_max: f32,
     /// Huber loss kappa for TD error (robust priority). 0.0 = disabled (raw L1).
     pub huber_kappa: f32,
-    /// D5: Whether to use NoisyNet exploration in the GPU kernel
-    pub use_noisy_nets: bool,
     /// D5: Initial noise std dev for NoisyNet (Rainbow DQN default: 0.5)
     pub noisy_sigma_init: f32,
-    /// D6: Whether to use C51 distributional RL
-    pub use_distributional: bool,
     /// D6: Number of atoms in C51 distribution (default: 51)
     pub num_atoms: i32,
     /// D6: Minimum value support for C51 (default: -50.0)
@@ -357,9 +353,7 @@ impl Default for ExperienceCollectorConfig {
             q_clip_min: -200.0,  // Reward v6: tighter than old -500 but covers v_range + safety margin
             q_clip_max: 200.0,
             huber_kappa: 0.0,
-            use_noisy_nets: true,   // Rainbow DQN: NoisyNet exploration always enabled
             noisy_sigma_init: 0.5,
-            use_distributional: true,  // Rainbow DQN: C51 distributional RL always enabled
             num_atoms: 51,          // C51 standard atom count
             // v_min/v_max: tight support for 16× atom resolution (delta_z=0.6 vs 9.6)
             v_min: -15.0,

crates/ml/src/trainers/dqn/trainer/training_loop.rs

@@ -1285,9 +1285,7 @@ impl DQNTrainer {
             } else {
                 0.0
             },
-            use_noisy_nets: true,
             noisy_sigma_init: self.hyperparams.noisy_sigma_init as f32,
-            use_distributional: true,
             num_atoms: self.hyperparams.num_atoms as i32,
             v_min: self.hyperparams.v_min as f32,
             v_max: self.hyperparams.v_max as f32,