[PyTorch Edge] Using Qnnpack in Quantized Softmax Op (#75799)
Summary:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/75799
Use Qnnpack's quantized softmax in quantized::softmax op when available
Test Plan:
From fbcode
```buck test caffe2/test:quantization -- test_qsoftmax```
# Benchmarking
(Naive Quantized from D35469257 v1, rest D34996486 v14)
|Shape|Fp32|Naive Quantized|Qnnpack Quantized|Qnnpack Quantized with Permute|
|(1, 5, 49, 49)|[6.6757](https://www.internalfb.com/intern/aibench/details/916894241135767)|[7.5981](https://www.internalfb.com/intern/aibench/details/504937774229694)|[1.5579](https://www.internalfb.com/intern/aibench/details/197716001861453)|[2.8446](https://www.internalfb.com/intern/aibench/details/59311708375203)|
|(1, 9, 16, 128)|[7.8485](https://www.internalfb.com/intern/aibench/details/135980349949180)|[9.0499](https://www.internalfb.com/intern/aibench/details/10150813869685)|[1.8865](https://www.internalfb.com/intern/aibench/details/58396904565184)|[3.5282](https://www.internalfb.com/intern/aibench/details/24583753477273)|
|(1, 5, 49, 64)|[7.0626](https://www.internalfb.com/intern/aibench/details/232201930202347)|[8.1091](https://www.internalfb.com/intern/aibench/details/57639118425406)|[1.801](https://www.internalfb.com/intern/aibench/details/656994017385942)|[3.2989](https://www.internalfb.com/intern/aibench/details/518979104130992)|
|(1, 3, 196, 64)|[16.4717](https://www.internalfb.com/intern/aibench/details/895795134460898)|[18.1987](https://www.internalfb.com/intern/aibench/details/909875420196348)|[3.5657](https://www.internalfb.com/intern/aibench/details/206864227381228)|[8.4519](https://www.internalfb.com/intern/aibench/details/84462467166362)|
|(1, 6, 49, 128)|[15.9872](https://www.internalfb.com/intern/aibench/details/417436371026264)|[17.4556](https://www.internalfb.com/intern/aibench/details/183113464145486)|[3.3912](https://www.internalfb.com/intern/aibench/details/616978041358188)|[8.019](https://www.internalfb.com/intern/aibench/details/849820562672950)|
|(1, 3, 196, 196)|[47.3636](https://www.internalfb.com/intern/aibench/details/633568439089073)|[52.0079](https://www.internalfb.com/intern/aibench/details/742080402804069)|[8.5009](https://www.internalfb.com/intern/aibench/details/685773806433926)|[13.5807](https://www.internalfb.com/intern/aibench/details/871998384861927)|
|(1, 6, 16, 64)|[4.0205](https://www.internalfb.com/intern/aibench/details/380419433454222)|[4.5973](https://www.internalfb.com/intern/aibench/details/923432861470595)|[1.0569](https://www.internalfb.com/intern/aibench/details/176718883676884)|[2.0519](https://www.internalfb.com/intern/aibench/details/303780226597723)|
|(1, 6, 16, 16)|[1.8299](https://www.internalfb.com/intern/aibench/details/599824935422385)|[2.3109](https://www.internalfb.com/intern/aibench/details/669753943440643)|[0.808](https://www.internalfb.com/intern/aibench/details/956331973568963)|[1.6406](https://www.internalfb.com/intern/aibench/details/924887465284668)|
|(1, 9, 16, 49)|[4.5134](https://www.internalfb.com/intern/aibench/details/946070183169117)|[5.2282](https://www.internalfb.com/intern/aibench/details/623403709385332)|[2.8195](https://www.internalfb.com/intern/aibench/details/635876531473203)|[2.2251](https://www.internalfb.com/intern/aibench/details/507256033953952)|
|(1, 6, 49, 196)|[23.9811](https://www.internalfb.com/intern/aibench/details/605021113223196)|[26.2834](https://www.internalfb.com/intern/aibench/details/991778071254930)|[4.5338](https://www.internalfb.com/intern/aibench/details/626603993142478)|[9.3877](https://www.internalfb.com/intern/aibench/details/962263658487065)|
*table made with https://www.internalfb.com/intern/anp/view/?id=1714217&revision_id=686803042569716*
Reviewed By: kimishpatel
Differential Revision: D34953197
fbshipit-source-id: 57418757fce17903583c04dffd51c886f9e1bc0e
(cherry picked from commit 8978222623f0cbacdb0373c405136ec94c035da6)