Add unit-aware gradients and corresponding documentation

[chaoming0625]

This pull request introduces the autograd module to the brainunit package, which includes functionality for automatic differentiation. The key changes involve adding new files for the autograd module, updating the __init__.py file to include the new module, and modifying existing methods for better compatibility.

New autograd module:

• brainunit/autograd/__init__.py: Added the autograd module with functions for automatic differentiation, such as value_and_grad, grad, vector_grad, jacobian, jacrev, jacfwd, and hessian.
• brainunit/autograd/_hessian.py: Implemented the hessian function, which computes the Hessian of a function with physical unit awareness.
• brainunit/autograd/_jacobian.py: Implemented jacrev, jacfwd, and jacobian functions for computing Jacobians with physical unit awareness.

Updates to existing files:

• brainunit/__init__.py: Imported the new autograd module and updated the __all__ list to include autograd. [1] [2]
• brainunit/_base.py: Modified the reshape method signature to accept a single shape argument instead of variadic arguments.

New tests:

• brainunit/autograd/_hessian_test.py: Added unit tests for the hessian function to verify its correctness with scalar, vector, and dictionary functions.
• brainunit/autograd/_jacobian_test.py: Added unit tests for jacrev and jacfwd functions to ensure they handle various cases correctly, including auxiliary outputs and multiple arguments.