Serializable cfg entries

[tensionhead]

As reported by @kajal5888, if a frontend parameter (e.g. foilim) is set via a non-serializable datatype (like numpy's int64) the standard spy.save fails because the cfg dictionary can't be dumped in to the accompanying json. Easiest to reproduce is sth like this:

my_data = ...  # some AnalogData
cfg = spy.StructDict()                                                                                                                           
cfg.foilim = [80, np.int64(120)]

spec = spy.freqanalysis(my_data, cfg)

spec.save('test')

this fails with

TypeError: Object of type int64 is not JSON serializable

How to Fix Idea

When updating the .cfg property of a Syncopy data object (for replayabilty) inside a frontend, it should be checked that all values are serializable, and if possible casted to appropriate types. So np.int64 -> int, np.float64 -> float and so on

[dfsp-spirit]

On the PR: This could actually be relevant in more places than just the foilim, maybe we want more general checks, as mentioned by @tensionhead.

Another question is whether we want to reject or try to automatically convert numpy datatypes as input in cases when they would go into the cfg. I think converting would be friendly, as using numpy datatypes as input seems a natural thing to do in some cases.

[tensionhead]

Yes, afaik all the numpy arrays we do accept in the frontend are of 'the sequence type', e.i. 1d-arrays. For those calling .tolist(() before casting to a serializable data type should be straightforward.

[tensionhead]

Got done with #407