New styles of event-driven computation operators

[chaoming0625]

This pull request includes several changes across multiple files to add new features, improve existing functionality, and update licensing information. The most important changes are summarized below:

New Features and Implementations:

• Added new implementations for the benchmark files COBA_2005.py and CUBA_2005.py, including detailed documentation, initialization of network parameters, and simulation functions. [1] [2]

Licensing Updates:

• Added Apache License 2.0 headers to multiple files to ensure proper licensing and compliance. [1] [2] [3]

Codebase Enhancements:

• Updated brainstate/event/__init__.py to include imports from the new _xla_custom_op module, and adjusted the __all__ list accordingly.
• Refactored brainstate/event/_csr.py to improve parameter naming and add support for new size types. Updated the CSRLinear class and its methods for better clarity and functionality. [1] [2] [3] [4] [5] [6] [7]

Testing and Benchmarking:

• Added new benchmarking scripts in brainstate/event/_fixed_probability_benchmark.py to measure performance improvements and acceleration ratios.
• Updated tests in brainstate/event/_fixed_probability_test.py to use the new FixedProb module path and ensure proper functionality. [1] [2]