create socket before starting the engine and pass socket into the vllm server on startup

[jberkhahn]

VLLM made a breaking change recently here: vllm-project/vllm#13113

But it looks like they've been doing stuff around passing that socket in for a while to avoid some stuff like race conditions? Figured I'd port over the functionality and get it on vllm version.

[codecov-commenter]

Codecov Report

All modified and coverable lines are covered by tests ✅

Project coverage is 63.46%. Comparing base (197f101) to head (a25711b).

Additional details and impacted files

@@            Coverage Diff             @@
##             main     #233      +/-   ##
==========================================
+ Coverage   63.26%   63.46%   +0.19%     
==========================================
  Files          28       28              
  Lines        1764     1768       +4     
  Branches      218      219       +1     
==========================================
+ Hits         1116     1122       +6     
+ Misses        542      541       -1     
+ Partials      106      105       -1     

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

[dtrifiro]

Thanks, looks good.