Ensure every instance of Simulation class has dataset, fix check_macro_cache

[anth-volk]

Fixes #262.

Previously, the Simulation class took an optional argument defaulted to None for a dataset parameter, which was used only when extending Simulation into classes like MicroSimulation and was always created with a default value of None if not provided. If a dataset was provided, it appears to have correctly integrated that into the Simulation instance. However, a dataset is not always provided, particularly by pe-core's test runner, which does not provide one.

This creates a confusing behavior whereby the dataset argument is always present, but an individual Simulation instance's self.dataset property is only created when dataset is not None. Furthermore, there are times when it appears self.dataset is created, but made equal to None. Furthermore, it's fairly easy to confuse the self.dataset instance attribute with the self.datasets class attribute, which is always present.

In order to disambiguate, this PR always instantiates a self.dataset instance attribute based upon the dataset arg, which continues to have a default value of None. Relying upon the expectation that self.dataset is always defined, it modifies the self.check_macro_cache method, but also adds a safety valve check to properly return False if self.dataset does not exist, just in case this PR missed some behavior somewhere.

[MaxGhenis]

Thanks Anthony. Merging to address the immediate needs. More motivation to consider how we can improve test coverage in core though.