Fix convnext to work with any custom input tensors

[Inquisitive-ME]

Convnext fails with custom inputs that are more than 1 layer, it looks like a bug. I added a test that highlights the problem and fails for all Convnext and then fixed the issue.

[codecov-commenter]

Codecov Report

All modified and coverable lines are covered by tests ✅

Project coverage is 84.87%. Comparing base (b0fc0a3) to head (6cbfb54).
Report is 3 commits behind head on master.

Additional details and impacted files

@@            Coverage Diff             @@
##           master   #20854      +/-   ##
==========================================
+ Coverage   82.25%   84.87%   +2.62%     
==========================================
  Files         559      559              
  Lines       52374    52388      +14     
  Branches     8096     8102       +6     
==========================================
+ Hits        43079    44464    +1385     
+ Misses       7305     5794    -1511     
- Partials     1990     2130     +140     

Flag	Coverage Δ
keras	84.69% <100.00%> (+2.62%)	⬆️
keras-jax	66.82% <100.00%> (+2.63%)	⬆️
keras-numpy	59.10% <0.00%> (+0.11%)	⬆️
keras-openvino	32.67% <0.00%> (+0.12%)	⬆️
keras-tensorflow	67.46% <100.00%> (+2.63%)	⬆️
keras-torch	66.78% <100.00%> (+2.62%)	⬆️
keras.applications	83.60% <100.00%> (?)
keras.applications-jax	83.60% <100.00%> (?)
keras.applications-numpy	22.81% <0.00%> (?)
keras.applications-openvino	22.81% <0.00%> (?)
keras.applications-tensorflow	83.60% <100.00%> (?)
keras.applications-torch	83.33% <100.00%> (?)

Flags with carried forward coverage won't be shown. Click here to find out more.

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

[fchollet]

Thanks for the fix! Please fix the code format (run sh shell/format.sh)

[fchollet]

LGTM, thank you