change softmax_lse correction of CP to FP32

[xrennvidia]

Description

• softmax_lse correction is in FP64 now, we can lower it to FP32.
• use log1p to be consistent with PR1401.

Type of change

• [ ] Documentation change (change only to the documentation, either a fix or a new content)
• [ ] Bug fix (non-breaking change which fixes an issue)
• [ ] New feature (non-breaking change which adds functionality)
• [ ] Breaking change (fix or feature that would cause existing functionality to not work as expected)
• [ ] Infra/Build change
• [x] Code refactoring

Checklist:

• [x] I have read and followed the contributing guidelines
• [x] The functionality is complete
• [x] I have commented my code, particularly in hard-to-understand areas
• [ ] I have made corresponding changes to the documentation
• [x] My changes generate no new warnings
• [ ] I have added tests that prove my fix is effective or that my feature works
• [x] New and existing unit tests pass locally with my changes

[xrennvidia]

/te-ci pytorch L1

[xrennvidia]

/te-ci pytorch L1

[xrennvidia]

/te-ci pytorch L1

[cyanguwa]

CI failures are not related to this PR.

log1p(x) provides more accuracy when x is close to 0.

Downcasting softmax_lse from double to float because cuDNN softmax_lse is in FP32 and combination of multiple copies of them over CP ranks should still stay in the float limits (see here).