Remove obsolete codes from PARAM

[TaekyungHeo]

Summary

Remove obsolete codes from PARAM. README will be updated by follow-up PRs after refactoring.

• Remove inference/compute/pt/pytorch_linear.py
• Remove train/compute/pt/*
• Remove dlrm
• Remove train/comms/pt/comms.py

Test Plan

$ grep "dlrm" -r          
./README.md:3. DLRM: tracks the `ext_dist` branch of DRLM benchmark use Facebook's DLRM benchmark (https://github.com/facebookresearch/dlrm). In short, PARAM fully relies on DLRM benchmark for end-to-end workload evaluation; with additional extensions as required for scale-out AI training platforms.
./train/comms/pt/README.md:The DLRM-Comms benchmark (`dlrm.py`) is similar to the open-source DLRM benchmark except it
./train/comms/pt/README.md:### DLRM-Comms benchmark (`dlrm.py`)
./train/comms/pt/README.md:mpirun -np <num-processes> -N <processes per node> --hostfile <file contains host list> ./dlrm.py \
./train/comms/pt/README.md:mpirun -np 16 -N 8 --hostfile ./hfile ./dlrm.py --master-ip $(head -n 1 ./hfile.txt) --mini-batch-size 32 \
./train/comms/pt/pytorch_dist_backend.py:        # Open-source extend_distributed.py can be found in https://github.com/facebookresearch/dlrm
./train/comms/pt/pytorch_dist_backend.py:        # TODO: this is a temporary workaround; need to unify the type of commsParams in comms and dlrm

$ find . -name 'pytorch_linear.py'

$ grep "pytorch_linear" -r

$ grep "compute.pt" -r        

$ grep commsCollBench -r             

$ grep MultilineFormatter -r