Update logic related to NestedModelField and fix several bugs

[ADR-007]

• Allow NestedModelField in ListField (any level. E.g.: [{a: {b: [{c: {d: 'e'}}]}}])
• Allow update/merge with None (bugfix?)
• Allow nested fields as model instances in __init__ method (bugfix? e.g. MyModel(nested=MyNested()))
• Allow update/merge with None in a deeply nested field (bugfix?)
• Wrap errors to add field path to the error message (e.g.: ... "a.b[0].c.d": field is required). Most likely, it will not break backward compatibility because there wasn't a base exception class, so most likely no one used particle exception in except statement
• Fix model.update (bug)
• Fix memory leak
• Set Meta.ignore_none_field=False by default. Most likely, it will not brake back-compatibility because True value was used to hide bugs in update. But... I think we can ignore None on create but not on update 🤔. It is not really clear how it should work. It probably should be "ignore default None, but save if it was set manually"
• Refactor some related code:
  • Add base exception class for model exceptions
  • Remove unused code
  • Remove conditions used to hide/resolve bugs for some cases
  • Move all the serialization/deserialization logic to Model. So, now there is only one place for it, which allows handling nested fields in a more convenient way, and overrides this behavior in NeatedModelField.get_value method.

Do I have a chance to have this merged, or is it too big one? Not sure how to do it in smaller PRs btw. I would like to use this code in a project I'm currently working on. So, I'll test it on real data as soon I'll know I have a chance to have it merged :)

[ADR-007]

There is a couple of more things I want to add to the project, but after this PR :)