Fix class instantiations

[BalzaniEdoardo]

Some functions (like np.split and others) accidentally dropped metadata. The issue was caused by the way the output TsdFrame was constructed by calling __class__ without explicitly routing the metadata of the original object.

Here I added a new abstract method of _Base named _define_instance for now, that should make sure that the appropriate properties are passed down to the newly defined object.

The method allow overwriting any parameter by passing new kwargs, if a kwarg is not passed, the parameter is grabbed from the attributes of self. I replaced all __class__(...) with _define_instance(...) everywhere in the code.

Testing

• _define_instance: test that it behaves as expected for all time_series.
• Class re-instantiaiton logic: check that all classes instantiation pass through _define_instance instead of __class__ directly. The test iterates over all pynapple funcitons, _Base methods and _Base subclasses methods, making sure that __class__ is not called directly.

[codecov]

Codecov Report

Attention: Patch coverage is 87.30159% with 8 lines in your changes missing coverage. Please review.

Files with missing lines	Patch %	Lines
pynapple/core/base_class.py	84.00%	2 Missing and 2 partials ⚠️
pynapple/core/time_series.py	88.57%	2 Missing and 2 partials ⚠️

Files with missing lines	Coverage Δ
pynapple/core/utils.py	84.09% <100.00%> (-0.18%)	⬇️
pynapple/process/filtering.py	94.44% <100.00%> (-0.16%)	⬇️
pynapple/core/base_class.py	97.60% <84.00%> (-1.51%)	⬇️
pynapple/core/time_series.py	89.77% <88.57%> (+0.21%)	⬆️

... and 2 files with indirect coverage changes