fix: softmax x86 miss tail data feat: use rcp to replace 1/x

[futz12]

fix: softmax x86 miss tail
feat: use rcp to replace 1/x

[codecov-commenter]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 95.89%. Comparing base (adee6a0) to head (ce52997).
⚠️ Report is 7 commits behind head on master.

Additional details and impacted files

@@            Coverage Diff             @@
##           master    #6368      +/-   ##
==========================================
- Coverage   95.89%   95.89%   -0.01%     
==========================================
  Files         840      840              
  Lines      265942   265941       -1     
==========================================
- Hits       255022   255021       -1     
  Misses      10920    10920              

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.

[Copilot]

Pull Request Overview

This PR fixes a bug in the x86 softmax implementation where tail data was incorrectly processed, and improves performance by replacing division operations with optimized reciprocal calculations using Newton-Raphson refinement.

Key changes:

• Fixed calculation of nn_size to properly handle tail elements using ceiling division: (size + sizen - 1) / sizen
• Introduced optimized reciprocal functions (_mm_rcp_nr_ps, _mm256_rcp_nr_ps, _mm512_rcp_nr_ps) that use Newton-Raphson refinement for better accuracy than raw rcp approximations
• Replaced all _mm*_div_ps division operations with the new reciprocal functions throughout the softmax implementations

---

_{Tip: Customize your code reviews with copilot-instructions.md. Create the file or learn how to get started.}

[github-actions]

The binary size change of libncnn.so (bytes)

architecture	base size	pr size	difference
x86_64	15190272	15190272	0 😘
armhf	6196764	6196716	-48 😘
aarch64	9522968	9523440	+472 ⚠️

[nihui]

Thanks for your contribution !