Add scan op

[james77777778]

Related to #19519

Originally, tf.scan stacks the output of the f as the final output and doesn't support the separated carry throughout the loop iterations.
I've reimplemented scan for tensorflow backend to align with the behavior of jax.lax.scan.

However, the current implementation requires that the output (y) of f must have the same shape/dtype as carry. This is not required in other backends.

[codecov-commenter]

Codecov Report

Attention: Patch coverage is 96.62162% with 5 lines in your changes are missing coverage. Please review.

Project coverage is 78.49%. Comparing base (10c27c0) to head (fbab23e).

Files	Patch %	Lines
keras/src/backend/tensorflow/core.py	92.30%	2 Missing and 2 partials ⚠️
keras/api/_tf_keras/keras/ops/__init__.py	0.00%	1 Missing ⚠️

Additional details and impacted files

@@            Coverage Diff             @@
##           master   #19681      +/-   ##
==========================================
+ Coverage   78.42%   78.49%   +0.07%     
==========================================
  Files         498      498              
  Lines       45551    45699     +148     
  Branches     8394     8448      +54     
==========================================
+ Hits        35723    35872     +149     
+ Misses       8094     8093       -1     
  Partials     1734     1734              

Flag	Coverage Δ
keras	78.34% <96.62%> (+0.07%)	⬆️
keras-jax	61.94% <19.59%> (-0.13%)	⬇️
keras-numpy	56.30% <39.86%> (-0.05%)	⬇️
keras-tensorflow	63.38% <48.64%> (-0.05%)	⬇️
keras-torch	62.01% <39.18%> (-0.08%)	⬇️

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]

LGTM -- great job!