Properly compare floats in fully-connected-sparse-operator-tester (#82768)
Similar to https://github.com/pytorch/pytorch/pull/82688
Example of current failure:
```
xplat/caffe2/aten/src/ATen/native/quantized/cpu/qnnpack/test/fully-connected-sparse-operator-tester.h:582
Expected equality of these values:
output_dynamic[i * outputChannels() + c]
Which is: -3064.26
accumulators_float[i * outputChannels() + c]
Which is: -3064.26
at 0, 1: reference = -3064.2568359375, optimized = -3064.256591796875
```
Pull Request resolved: https://github.com/pytorch/pytorch/pull/82768
Approved by: https://github.com/seemethere