whl centric workflow for "remote push"

[andy-neuma]

SUMMARY:

• moves "nm-vllm" to whl centric workflows. this means that testing is conducted on a fresh instance via installing "nm-vllm" whl into a venv.
• update build action to also generate whl and source distributions.
• update testing action to work via whl install.
• updates callable build workflow to build, generate whl, and source distribution. artifacts are uploaded to GHA artifact store.
• updates callable test workflow to install whl and run tests.
• patch code coverage to work with "forked" processes

NOTES: currently we are building on a GPU instance, but technically this isn't a requirement. the current non-gpu instance we have in our runner group doesn't have enough memory. this results on builds sporadically failing, so "remote push" is set up to run on T4's.

TEST PLAN:
runs on remote push

[dhuangnm]

The tarball generation is covered in my PR: #138

Maybe we can merge the two or separate the tarball part out in its own PR.

[andy-neuma]

@dhuangnm - as of this diff, we will always be building whl and source distributions.

[varun-sundar-rabindranath]

LGTM !