Add JSON conversion warnings, store save-type metadata and use ns file timestamps (v0.8.2)

[kyrylo-gr]

Motivation (#6)

• Improve visibility when lists are coerced to JSON strings during save and allow controlling warning behavior.
• Record how values were stored (e.g. JSON string, H5 dataset, serialized function) and surface that in the object representation.
• Increase file-modification precision and robustness by using nanosecond timestamps and file size to detect external changes.

Description

• Added a JSON-conversion warning helper _warn_json_conversion_if_needed and extended transform_not_dict_on_save to accept json_conversion_mode and key so conversions can warn in modes all, auto, or be muted; recursive calls mute warnings.
• Extended save_sub_dict and save_dict to accept json_conversion_mode and a storage_type dict to collect metadata about how each key was stored, and changed save_dict to return nanosecond-resolution modification time (int) instead of float mtime.
• Tracked file st_size and st_mtime_ns in DH5, updated loading/saving logic to use os.stat(...).st_mtime_ns and st_size, and updated pull_available to compare both mtime and size.
• Stored per-key saved-type metadata in DH5._saved_storage_type and appended saved-type postfixes to the repr via output_dict_structure additional info.
• Added validation for constructor parameter json_conversion_mode and defaulted it to "auto" in DH5.
• Minor refactor/formatting in dict_structure.py and ensured warnings/logging import in data_transformation.py.

Testing

• Ran the unit tests in tests/dh5_class_tests/main_test.py, including newly added test_repr_contains_saved_type_after_save and test_json_conversion_warning_modes, and they passed under the unittest run.
• Existing API-level tests (save/open/representation) were executed and succeeded.