Clean up training configs

[lchu6]

a list of minors around training configs that should be fixed later.

1. this should really be > num_step rather than == num_step, as our batch_idx starts from 1 rather than 0. The current implementation would skip the required last step causing last checkpoint won't be written.
2. we should remove cfg. here
3. remove sharding_group_size in training configs, as we dropped the support for both SSDP and TP in the open source version.
4. re-order training configs to a more meaning grouping.