Reduce dependency install time and decrease docker image size

[mreso]

Description

This PR changes the order of the installation of torchserve dependencies from "requirements.txt -> torch" into "torch -> requirements.txt". This speeds up the installation of the dependencies and reduces the docker CPU image size from 4.98GB to 2GB and GPU from 11.1GB to 8.4GB.

Fixes #2358

Type of change

Please delete options that are not relevant.

• Bug fix (non-breaking change which fixes an issue)

Feature/Issue validation/testing

Please describe the Unit or Integration tests that you ran to verify your changes and relevant result summary. Provide instructions so it can be reproduced.
Please also list any relevant details for your test configuration.

• images size before:
  CPU:
  
  GPU:
  
  images size after:
  CPU:
  
  GPU:
  

• docker run --rm -it -p 8080:8080 -p 8081:8081 -p 8082:8082 -p 7070:7070 -p 7071:7071 2db5237a38ea
  

Checklist:

• Did you have fun?
• Have you added tests that prove your fix is effective or that this feature works?

[codecov]

Codecov Report

Merging #2364 (86a8f05) into master (4f85477) will decrease coverage by 0.11%.
The diff coverage is n/a.

❗ Current head 86a8f05 differs from pull request most recent head d248555. Consider uploading reports for the commit d248555 to get more accurate results

@@            Coverage Diff             @@
##           master    #2364      +/-   ##
==========================================
- Coverage   70.69%   70.58%   -0.11%     
==========================================
  Files          78       78              
  Lines        3638     3638              
  Branches       57       57              
==========================================
- Hits         2572     2568       -4     
- Misses       1063     1067       +4     
  Partials        3        3              

see 2 files with indirect coverage changes

📣 We’re building smart automated test selection to slash your CI/CD build times. Learn more

[msaroufim]

Well actually in this case we're not guaranteed to have the torch version that shows up in our requirements as the one that gets installed

[mreso]

True. For this we need to freeze all version numbers of our dependencies which we should do in any case.

[mreso]

@agunapal This fixes the nightly docker build: https://github.com/pytorch/serve/actions/runs/5085378605