[MLF] Allow for computing prune quantile thresholds on absolute value of indicators in distributed-inference-compatible embedding LUT pruning (#46789)
Summary:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/46789
1. Now `SelfBinningHistogram` can calculate the binning histogram using the absolute values from the given an array of values.
2. Update the invocation of `SelfBinningHistogram` in `post_training_prune`.
Test Plan:
1. [buck test caffe2/caffe2/python/operator_test:self_binning_histogram_test](https://www.internalfb.com/intern/testinfra/testconsole/testrun/6473924488326108/)
2. [buck test dper3/dper3_backend/delivery/tests:post_training_prune_test](https://www.internalfb.com/intern/testinfra/testconsole/testrun/2251799854023163/)
Reviewed By: hwangjeff
Differential Revision: D24494097
fbshipit-source-id: 95e47137b25746e686ef9baa9409560af5d58fc1
