ART-Optimized Megatron (AOM): 12x higher training throughput

Introduced ART-Optimized Megatron (AOM) as a new backend for the ART training library, achieving 12x higher training step throughput. Implemented shared prefix training using PyTorch's FlexAttention and custom Gated DeltaNet (GDN) layer modifications to handle non-linear tree layouts, significantly increasing trajectory density. Optimized packed sequence length for maximum GPU throughput and tuned parallelism configurations for improved trajectories per second. This builds upon existing distributed training resiliency efforts by directly enhancing the training throughput component.