feat(meta,jepa): expose GPU features through meta-crates; wire JEPA cluster plan and real shard loading

Meta-crates (Phase 2):
- rtx-core / rtx-training / rtx-inference-stack gain cuda and metal
  features threading into their sub-crates; GPU was previously
  unreachable through the user-facing bundles.
- rtx-training restores rtx-distributed (the hpc-channels blocker is
  gone) so the advertised DistributedTransformerTrainer resolves; drops
  the unused rtx-runtime dep.
- rtx-transformers drops unused rtx-backend/rtx-backend-cpu deps
  (stale comment referenced a teacher that never used them).

Never-compiled CUDA paths fixed (surfaced by the new feature wiring,
verified on RTX 5060 Ti / CUDA 13.1):
- rtx-compress build.rs: missing Path/Command/fs imports.
- rtx-flash-attention flash_decode_forward: reborrow &mut kernel args.
- rtx-transformers: rope kernel include path, cudarc 0.18 Arc<CudaModule>,
  PushKernelArg imports in jepa_gpu, edition-2024 ref patterns.
- rtx-memory: full cudarc 0.18 port (CudaContext, stream-based alloc,
  DevicePtr accessors, error enum formatting) across gpu_pinning,
  gpu_transfer, gpu_real, gpu_allocator/arena, gpu_tests.

JEPA (Phase 3):
- JepaRunConfig::apply_cluster_plan consumes ClusterTrainingPlan
  (batch size, TP/DP, world size, total steps) so jepa_cluster is no
  longer standalone dead config; ViTSizeStr::approx_params_m feeds
  JepaParallelConfig::for_model_and_cluster.
- WebDatasetShard::load reads real .tar shards from disk via the
  existing parser (gzip rejected explicitly); to_in_memory documented
  as synthetic/test-only.
- New image-decode feature actually defines the dep for the previously
  unreachable cfg(feature = "image-decode") JPEG/PNG decode path.

Co-Authored-By: Claude Fable 5 <[email protected]>

crates/meta/rtx-training/src/lib.rs

@@ -15,8 +15,8 @@ pub use rtx_tensor as tensor;
 
 // Training
 pub use rtx_transformers as transformers;
-// pub use rtx_distributed as distributed;  // Temporarily disabled - missing hpc-channels
 pub use rtx_compress as compress;
+pub use rtx_distributed as distributed;
 pub use rtx_flash_attention as flash_attention;
 pub use rtx_rl as rl;
 
@@ -25,8 +25,7 @@ pub mod prelude {
     pub use rtx_tensor::{Device, Tensor};
     // Use the correct exports from rtx_transformers
     pub use rtx_transformers::architectures::{TransformerArchitecture, TransformerConfig};
-    // rtx_distributed temporarily disabled - missing hpc-channels
-    // pub use rtx_distributed::distributed_transformer_trainer::DistributedTransformerTrainer;
+    pub use rtx_distributed::distributed_transformer_trainer::DistributedTransformerTrainer;
     pub use rtx_flash_attention::FlashAttention;
     // Use the correct exports from rtx_rl
     pub use rtx_rl::actor_learner::ActorLearner;