[Fix] DDP init with device id

[felixdittrich92]

This pull request improves distributed training reliability and device management for all detection, recognition, layout, and table training scripts. The main changes ensure that PyTorch's distributed process group is correctly initialized with device information and that CUDA-aware barriers are used for synchronization, which helps prevent deadlocks and device mismatch errors.

Distributed Training Initialization:

• All train.py scripts (detection, recognition, layout, table) now initialize the process group with the device_id argument, ensuring each process is explicitly associated with the correct CUDA device. [1] [2] [3] [4]

Barrier Synchronization Improvements:

• The barrier_download context manager in all ddp_utils.py modules (detection, recognition, layout, table) now uses a helper function to call dist.barrier with device_ids when using the NCCL backend and CUDA, ensuring correct device synchronization and avoiding potential deadlocks.

These changes collectively improve robustness and compatibility for multi-GPU distributed training across all relevant modules.

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 97.06%. Comparing base (d7a65fa) to head (9461c8c).
⚠️ Report is 1 commits behind head on main.

Additional details and impacted files

@@            Coverage Diff             @@
##             main    #2136      +/-   ##
==========================================
+ Coverage   97.01%   97.06%   +0.04%     
==========================================
  Files         169      169              
  Lines       10000    10000              
==========================================
+ Hits         9701     9706       +5     
+ Misses        299      294       -5     

Flag	Coverage Δ
unittests	97.06% <ø> (+0.04%)	⬆️

Flags with carried forward coverage won't be shown. Click here to find out more.

☔ View full report in Codecov by Harness.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.