Merge pull request alibaba#1772 from alibaba/feature/sync

[MNN:Sync] Sync internal gitlab

[MNN:Bugfix] Fix compile bug for gnu of arm82 /bf16

Fix bug for not change input for nhwc -> nc4hw4

Fix bug for dim = 0's shape compute

Fix bug for Convert for HardSigmoid without default value for beta

Synchronize internal github for version 1.2.0 (alibaba#1518)

Merge pull request alibaba#1395 from WillTao-RD/master

fix opencl runtime error of MSVC; add 'clGetMemObjectInfo' wrapper

Merge pull request alibaba#1454 from alibaba/feature/sync

Feature/sync

Merge pull request alibaba#1437 from alibaba/feature/sync

[MNN:Sync] Sync internal Gitlab ，CPU FP16 / BF16 Opt , TensorArray bugfix, Full-Quantization

Merge pull request alibaba#1406 from alibaba/feature/syncopencl

[MNN:Sync] Sync opencl from Internal Gitlab, support Buffer and more auto-turning mode