[PERFORMANCE] Fix Packing thread regression in code

[Qubitium]

Reason for PR:

• Packing has unfixed thread regression on env with lots of cpu cores. The more cores you have, the slower your packing time becomes. Non-linear regression.
• Previous fix was simple warning on injecting env vars. Force warn users to reduce threads or packing may regress ~100x #628
• PR fixes this issue, not the core cause, by limiting packing() context to use 1 thread. The "pack" procedure is extremely slow #439

However, this will need a new pkg depend: https://github.com/joblib/threadpoolctl

Python helpers to limit the number of threads used in native libraries that handle their own internal threadpool (BLAS and OpenMP implementations)

ref: #640 (comment)

Tests Passed:

• test_quantization
• test_serialization

[Qubitium]

No need to merge this if #640 is merged first as that contains everything here and more.