FEAT-#4527: Add Modin logging to AxisPartition and BlockPartition

[arunjose696]

What do these changes do?

In #4527 it was mentioned to add modin logging to different layers, Among these logging support has already been added to Partition Manager, The current PR adds logging to Block and Axis partitions.

For logging in remote function it would be recommended to use external tools specific to backends for example Ray timeline for ray backend , UNIDIST_MPI_LOG for unidist etc.

• first commit message and PR title follow format outlined here

  NOTE: If you edit the PR title to match this format, you need to add another commit (even if it's empty) or amend your last commit for the CI job that checks the PR title to pick up the new PR title.

• passes flake8 modin/ asv_bench/benchmarks scripts/doc_checker.py
• passes black --check modin/ asv_bench/benchmarks scripts/doc_checker.py
• signed commit with git commit -s
• Resolves Add Modin logging to engine layer #4527?
• tests added and passing
• module layout described at docs/development/architecture.rst is up-to-date

[YarShev]

Since we use the methods of this class inside remote functions, how does this addition of ClassLogger here affect logging for remote functions?

[arunjose696]

As I understand BaseDataframeAxisPartition's methods would call the remote functions, I could not see the other way around where methods of BaseDataframeAxisPartition are called in remote functions, I could see data of the partitions are send to remote functions and operations are performed on them, Is there anyplace where partitions would be directly passed to remote functions?

When running remote functions from the methods of BaseDataframeAxisPartition the class logger would work normally as it would log BaseDataframeAxisPartition methods and leave out logging of remote functions.

[YarShev]

BaseDataframeAxisPartition's methods do not call the remote functions. Instead, engine specific virtual partitions call remote functions inside of which BaseDataframeAxisPartition's methods get called. Here is an example where we pass an object reference to the method of BaseDataframeAxisPartition and object references to data partitions in the remote function.

modin/modin/core/execution/ray/implementations/pandas_on_ray/partitioning/virtual_partition.py

Lines 176 to 192 in 93b4e2a

	return _deploy_ray_func.options(
	num_returns=(num_splits if lengths is None else len(lengths))
	* (1 + cls._PARTITIONS_METADATA_LEN),
	**({"max_retries": max_retries} if max_retries is not None else {}),
	).remote(
	cls._get_deploy_axis_func(),
	*f_args,
	num_splits,
	maintain_partitioning,
	*partitions,
	axis=axis,
	f_to_deploy=func,
	f_len_args=len(f_args),
	f_kwargs=f_kwargs,
	manual_partition=manual_partition,
	lengths=lengths,
	return_generator=True,

Given that, I wonder if this logging works inside remote functions?

[arunjose696]

Got it Thanks

When we enable the logger with LogMode.enable() in modin code, the logger gets enabled only on the main process, The logger remains disabled in workers so the remote functions execute the function without logging so info about the function wouldn't be logged from remote function.

[YarShev]

Why should we disable logging for this method?

[arunjose696]

Logging requires modules like sys, which might already have been cleaned up by the time del is invoked(As we couldnt be sure when del would be invoked). This leads to ModuleNotFoundError. So disabling logging of del