make the project installable with pip

[rsxdalv]

pip install git+https://github.com/rsxdalv/Kimi-Audio@package

I also had to add a init.py to GLM-4-Voice so I forked it and only left the speech_tokenizer folder in the fork.

Since my intent is to use it with PyTorch 2.6.0, I loosened the requirements in setup.py.

Some debugging might still be required.

[MoyanZitto]

Thank you for your interest in our project. This PR has two aspects that we need to further verify, so it will not be processed immediately:

• We may need to conduct a complete validation of the dependency package versions and confirm their correctness.
• We should ensure that GLM4-Voice is installed from an official source. I believe there is a way to achieve this.