unroll n-tile x16 for gemm x86 avx512

[nihui]

• opt 8x16
• opt 4x16

[tencent-adm]

Thank you for your submission, we really appreciate it. Like many open source projects, we ask that you sign our Contributor License Agreement before we can accept your contribution.
_{You have signed the CLA already but the status is still pending? Let us recheck it.}

[codecov-commenter]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 95.90%. Comparing base (82af4a1) to head (6b709bf).

Additional details and impacted files

@@            Coverage Diff             @@
##           master    #6434      +/-   ##
==========================================
- Coverage   95.91%   95.90%   -0.02%     
==========================================
  Files         844      844              
  Lines      267021   265852    -1169     
==========================================
- Hits       256122   254955    -1167     
+ Misses      10899    10897       -2     

☔ 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.

[github-actions]

The binary size change of libncnn.so (bytes)

architecture	base size	pr size	difference
x86_64	15328640	15316352	-12288 😘
armhf	6229904	6229904	0 😘
aarch64	9527568	9527568	0 😘

[nihui]

r9-9950x single thread

$ ./benchllm-0 100 1 2 -1 0 233
loop_count = 100
num_threads = 1
powersave = 2
gpu_device = -1
cooling_down = 0
seqlen = 233
            minicpm4 (prefill)  min =  671.96  max =  678.33  avg =  673.73
            minicpm4  (decode)  min =   35.03  max =   36.12  avg =   35.23


$ ./benchllm 100 1 2 -1 0 233
loop_count = 100
num_threads = 1
powersave = 2
gpu_device = -1
cooling_down = 0
seqlen = 233
            minicpm4 (prefill)  min =  638.59  max =  646.15  avg =  640.96
            minicpm4  (decode)  min =   32.80  max =   34.75  avg =   33.17