extractor_dict_iterator for solving path detection in object kwargs

[h-mayorquin]

Extractor dicts (outputs of BaseExtractor.to_dict()) are nested structures where some arguments are paths.

For solving the problem in #3013 we need a way of filtering by the arguments that correspond to real paths (e.g. Path(argument).exists() is True) and only modify those.

The current approach is to use the recursive_path_modifier:
https://github.com/alejoe91/spikeinterface/blob/fa2ca6d7c804fc8ab3df93dbd463d9576bc7723b/src/spikeinterface/core/core_tools.py#L186

I guess at the beginning the purpose of the function was both iterating overt the dict and also modifying the values.
But then, some other functions appeared like get_path_list that use recursive_path_modifier to iterate over the extractor and accumulate path-like values.

This PR creates two functions that can accomplish the same as recursive_path_modifier but are decoupled:

1. First, it introduces extractor_dict_iterator that generates an iterator over all elements with three attributes: value, name and access path.
2. Second, a new set_value_in_extractor_dict function that uses the access path of the elements from the iterator to modify the value is also introduced.

With this logic, functions like find_recording_folders that only use the first part can avoid the part of recursive_path_modifier that modifies the values. Avoiding computational waste and simplifying intent. Moreover,, the problem in #3014 of having to check for path existence before or after a modification can be easily stated as a single line where elements are filtered instead of modifying the logic of recursive_path_modifier. In general it enables working with extractor dicts in the folowing way:

1. Collect items from the extractor dict
2. Filter to select some of them
3. Modify them

Note that this does not even need to be about paths. Any type of filter, reduce or concistency check and modification can be implemented with this.

I also modified the tests of make_relative_paths and make_absolute_paths to work on real paths so we can directly test the type of problem causes #3013

[alejoe91]

@h-mayorquin I love this! just some renaming suggestions

[alejoe91]

@h-mayorquin this should cloe #3014 right?

[h-mayorquin]

@alejoe91 did the correction and yes, this should close #3041

[alejoe91]

@h-mayorquin can you make the previous test function in test_core_tools (test_path_utils_functions) also to use pytest.skip?

[h-mayorquin]

Yes, done.

[alejoe91]

This looks good to me. Much much cleaner than the old presentation!

@samuelgarcia what do you think?

[samuelgarcia]

I guess at the beginning the purpose of the function was both iterating overt the dict and also modifying the values. But then, some other functions appeared like get_path_list that use recursive_path_modifier to iterate over the extractor and accumulate path-like values.

Excellent guess and analysis of the histical reason!
I wanted to do something similar to this PR this when impleneted _get_paths_list but was to lazy to refacor with the yield like you did.

Thanks a lot for this effort.

[samuelgarcia]

This is really smart to have the path both with key and index for list!!

[samuelgarcia]

very good idea!