fix bugs in truncated_normal

[charlielam0615]

Description

truncated_normal in brainpy.math.random calls jr.uniform to sample from a uniform distribution and then inverse the cdf to get truncated values.

jr.uniform samples from a closed interval. When lower or upper takes inf values and when jr.uniform sampled an extreme value, the inverse cdf would have inf values, which result in underflow values after clipping.

Fix: subtract and add a small value to minval and maxval arguments of jr.uniform.

How Has This Been Tested

Types of changes

• Bug fix (non-breaking change which fixes an issue)

Checklist

• Code follows the code style of this project.
• Changes follow the CONTRIBUTING guidelines.
• Update necessary documentation accordingly.
• Lint and tests pass locally with the changes.
• Check issues and pull requests first. You don't want to duplicate effort.

Other information

[chaoming0625]

Great PR. I have not aware of this issue!

[chaoming0625]

I recommend to use the following samll value:

eps = lax.nextafter(0., np.array(np.inf, dtype=dtype)))
out = jr.uniform(key, size, dtype, minval=2 * l - 1 + eps , maxval=2 * u - 1 - eps)

[chaoming0625]

Great work!