Apply backend.result_type to absolute, argmax, argmin, argsort, ceil, clip and dot

[james77777778]

Followed by #18482 and #18534

The corresponding unit tests have also been added.

I choose these ops that are more commonly used in the codebase.
In ceil, int64 input will become float64 output but I have changed it to backend.floatx().

The principles in my mind, which are derived from the previous discussion:

• match the type inference of JAX
• if the integer input becomes a floating point output, use backend.floatx()
• try to support all dtypes in all backends (even if it introduces some overhead for checking)

[codecov-commenter]

Codecov Report

Attention: 21 lines in your changes are missing coverage. Please review.

Comparison is base (fd5272b) 77.67% compared to head (c96ea77) 71.97%.

Additional details and impacted files

@@            Coverage Diff             @@
##           master   #18548      +/-   ##
==========================================
- Coverage   77.67%   71.97%   -5.71%     
==========================================
  Files         334      334              
  Lines       32302    32354      +52     
  Branches     6297     6308      +11     
==========================================
- Hits        25092    23287    -1805     
- Misses       5631     7578    +1947     
+ Partials     1579     1489      -90     

Flag	Coverage Δ
keras	71.90% <72.36%> (-5.68%)	⬇️
keras-jax	63.17% <23.68%> (-0.06%)	⬇️
keras-numpy	57.27% <39.47%> (-0.02%)	⬇️
keras-tensorflow	63.21% <43.42%> (-0.02%)	⬇️
keras-torch	?

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

Files	Coverage Δ
keras/backend/jax/numpy.py	99.17% <100.00%> (+0.28%)	⬆️
keras/backend/numpy/numpy.py	99.19% <100.00%> (+0.58%)	⬆️
keras/backend/tensorflow/numpy.py	95.74% <100.00%> (+0.35%)	⬆️
keras/ops/numpy.py	95.49% <91.66%> (+0.15%)	⬆️
keras/backend/torch/numpy.py	21.03% <0.00%> (-73.83%)	⬇️

... and 35 files with indirect coverage changes

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

[james77777778]

Some tests have failed and I believe the root cause is the new release of torch 2.1.0.
I'm working on the fix.

EDITED:
please see #18550

[fchollet]

Thanks for the PR. Great work!