Adds tests and fixes issues with CompileLoss true/pred/loss ordering

[gustavoeb]

This PR addresses the issue described here: #20346

• Firstly improved the existing test_list_loss_dict_data; the previously chosen values of a and b are such that no matter the order of the losses (mae, mse) the total loss value is always the same, hiding potential issues.
• Added a couple more tests to mimic what happens with functional models with multiple outputs, fitted with data coming from a dictionary.
• Keeps dicts in their existing order, or output_name order, by casting them to OrderDict to circumvent optree’s sorting behaviour

[codecov-commenter]

Codecov Report

Attention: Patch coverage is 85.71429% with 2 lines in your changes missing coverage. Please review.

Project coverage is 57.43%. Comparing base (3d32481) to head (fb04f63).

Files with missing lines	Patch %	Lines
keras/src/trainers/compile_utils.py	85.71%	2 Missing ⚠️

❗ There is a different number of reports uploaded between BASE (3d32481) and HEAD (fb04f63). Click for more details.

HEAD has 6 uploads less than BASE

Flag	BASE (3d32481)	HEAD (fb04f63)
keras	4	1
keras-torch	1	0
keras-tensorflow	1	0
keras-jax	1	0

Additional details and impacted files

@@             Coverage Diff             @@
##           master   #20362       +/-   ##
===========================================
- Coverage   78.87%   57.43%   -21.45%     
===========================================
  Files         512      512               
  Lines       49266    49280       +14     
  Branches     7953     7958        +5     
===========================================
- Hits        38861    28302    -10559     
- Misses       8546    19172    +10626     
+ Partials     1859     1806       -53     

Flag	Coverage Δ
keras	57.43% <85.71%> (-21.32%)	⬇️
keras-jax	?
keras-numpy	57.43% <85.71%> (+<0.01%)	⬆️
keras-tensorflow	?
keras-torch	?

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.