ARM: layernorm neon/fp16s/fp16sa/bf16s

[EdVince]

完成了layernorm的arm部分：

1. neon:pack1,pack4
2. fp16s:pack1,pack4,pack8
3. fp16sa:pack1,pack4,pack8
4. bf16s:pack1,pack4

(PS:犀牛鸟计划的工作)

[tencent-adm]

All committers have signed the CLA.

[codecov-commenter]

Codecov Report

Merging #4102 (2a125c7) into master (00c08d7) will decrease coverage by 14.45%. The diff coverage is n/a.

@@             Coverage Diff             @@
##           master    #4102       +/-   ##
===========================================
- Coverage   94.43%   79.97%   -14.46%     
===========================================
  Files         748      372      -376     
  Lines      179005    81508    -97497     
===========================================
- Hits       169047    65190   -103857     
- Misses       9958    16318     +6360     

Impacted Files	Coverage Δ
src/layer/x86/convolution_pack8.h	0.00% <0.00%> (-100.00%)	:arrow_down:
src/layer/x86/deconvolution_pack8.h	0.00% <0.00%> (-100.00%)	:arrow_down:
src/layer/x86/interp_bicubic_pack8.h	0.00% <0.00%> (-100.00%)	:arrow_down:
src/layer/x86/convolution_1x1_pack8.h	0.00% <0.00%> (-100.00%)	:arrow_down:
src/layer/x86/convolution_2x2_pack8.h	0.00% <0.00%> (-100.00%)	:arrow_down:
src/layer/x86/interp_bilinear_pack8.h	0.00% <0.00%> (-100.00%)	:arrow_down:
src/layer/x86/convolution_1x1_pack4to8.h	0.00% <0.00%> (-100.00%)	:arrow_down:
src/layer/x86/convolution_sgemm_pack4to8.h	0.00% <0.00%> (-100.00%)	:arrow_down:
src/layer/x86/convolution_sgemm_pack8.h	0.00% <0.00%> (-99.49%)	:arrow_down:
src/layer/x86/convolution_pack4to8.h	0.00% <0.00%> (-97.96%)	:arrow_down:
... and 706 more

Help us with your feedback. Take ten seconds to tell us how you rate us. Have a feature suggestion? Share it here.

[tpoisonooo]

没加 tests 和 benchmark.

[EdVince]

需要额外加吗？现在的test应该都够了吧。

[EdVince]

中文注释都去掉了，up有空看看代码吗😊 @nihui