Swap to use more portable types than pickles

[ljwolf]

This swaps to a serialized types.SimpleNamespace, rather than a pickle to compare tests against.

This is necessary for a few reasons.
First, when you pickle your object, python will execute its methods in the new context you're testing in. This means that its references to mgwr are what you're trying to test. So, when you run stuff like MGWR.local_multicollinearity(), it's running in the current, under testing mgwr package, not the cached result from a previous known-good mgwr run.
Second, unpickling the mgwr object requires that mgwr be installed. This assumption isn't tenable for downstream packaging in pysal.

What I've done is unpack the object using its __dict__ attribute and stuff it all inside of a types.SimpleNamespace. This is kind of like a dict with dot access, so I can write it out with whatever we need. And, it only depends on the standard library and numpy to initialize, not the mgwr package.

To make this happen, though, you actually have to run some methods in order to cache their results. I've done this, and stored their results in the SimpleNamespace. Now our comparisons compare the result.method() to the cached result of method().

[TaylorOshan]

This makes a lot of sense, looks good to me.

[TaylorOshan]

looks good!