Add model manager that automatically manage model across processes (#37113)

* Add model manager that automatically manage model across processes

* Add pydoc and move gpu detection to start

* Add comments and helper function to make it easier to understand the code and cleanup some code logics

* Add TODO for threading

* Remove tracked model proxy and have model manager store tags instead of model instance

* Fix import order

* Clean up and logs

* Added timeout for waiting too long on model acquire

* Throw error if timeout

* Add test for timeout and adjust

* Update sdks/python/apache_beam/ml/inference/model_manager.py

Co-authored-by: gemini-code-assist[bot] <176961590+gemini-code-assist[bot]@users.noreply.github.com>

* Update sdks/python/apache_beam/ml/inference/model_manager.py

Co-authored-by: gemini-code-assist[bot] <176961590+gemini-code-assist[bot]@users.noreply.github.com>

* Gemini clean up

* Update sdks/python/apache_beam/ml/inference/model_manager_test.py

Co-authored-by: gemini-code-assist[bot] <176961590+gemini-code-assist[bot]@users.noreply.github.com>

* Update GPU monitor test

* Format

* Cleanup upating is_unkown logic

* Try to fix flake

* Fix import order

* Fix random seed to avoid flake

* Fix identation

* Try fixing doc again

---------

Co-authored-by: gemini-code-assist[bot] <176961590+gemini-code-assist[bot]@users.noreply.github.com>

Author: AMOOOMA