Fix dimensionality issue when binary classification outputs 1D instead of 2D tensor

[XianzheMa]

For binary classification (e.g. DLRM), there are two problems:

1. The model output no longer has shape (batch_size, num_classes). The size is not even (batch_size, 1) as the second dimension, 1, is squeezed.
2. The model output is not passed through a softmax layer but a sigmoid layer.

This PR adapts the existing downsamplers to the binary classfication.

[codecov]

Codecov Report

Attention: Patch coverage is 96.77419% with 1 line in your changes missing coverage. Please review.

Project coverage is 85.62%. Comparing base (4e1550d) to head (eabe7de).
Report is 1 commits behind head on main.

Files with missing lines	Patch %	Lines
...mplers/remote_uncertainty_downsampling_strategy.py	95.23%	1 Missing ⚠️

Additional details and impacted files

@@            Coverage Diff             @@
##             main     #609      +/-   ##
==========================================
+ Coverage   85.60%   85.62%   +0.02%     
==========================================
  Files         258      258              
  Lines       11311    11327      +16     
==========================================
+ Hits         9683     9699      +16     
  Misses       1628     1628              

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

[MaxiBoether]

Thank you! This mostly looks good, I had a couple of clarification comments and suggestions.

[MaxiBoether]

Thank you, Xianzhe! This LGTM!