Fix ep deployment issues

[CUHKSZzxy]

Modifications

1. Expose deepep env var

Default deepep buffer num sms will raise the following errors on H200 multi-nodes. Therefore, we expose this environment variable to users for configuration. A feasible value on H200 is DEEPEP_BUFFER_NUM_SMS=16.

csrc/kernels/internode.cu:386, condition: ibgda_get_state()->num_rc_per_pe == num_channels or ibgda_get_state()->num_rc_per_pe >= num_sms

This is a known issue in deepep

• By default, in megatron, condition fais on ibgda_get_state()->num_rc_per_pe == num_channels || ibgda_get_state()->num_rc_per_pe >= num_sms  deepseek-ai/DeepEP#226

2. Fix deepgemm

• Pin deepgemm to an older version (1.0.0+03d0be3), for dlblas compatibility.
• For this specific deepgemm commit, we add its jit compile dependencies.

TODO

• Test and try the DeepEP stable release v1.2.1
• Test and try the DeepGEMM stable release