Add ProteinStructureStore

[AntonOresten]

This PR implements the Deserializer type (see #9). It's currently a lazy AbstractDict{String,ProteinStructure}, and deviates slightly from the original suggestion in that you iterate through the structures using values(deserializer), defined implicitly by explicitly defining Base.keys and Base.iterate.

As mentioned in #9, I'm also considering some kind of Serializer type for appending structures to files, but this can already be done with serialize(path, structures; mode="cw"). Still, the current interface provides no way to e.g. delete an entry without importing HDF5. A Serializer type could implement Base.delete!(serializer, key) method -- or should serializers strictly write, in the same way deserializers strictly read?

Should also reconcile the overlap between this and my make shift ProteinDataset type from #9. The deserialize function could be implemented as collect(values(Deserializer(filename))).

The commit also includes some minor package reformatting.

[bicycle1885]

Thank you! This looks good if you could add some tests to prevent regression in the future.

[AntonOresten]

I've replaced Deserializer and ProteinDataset with a more comprehensive ProteinStructureStore. ProteinDataset was really just a glorified Dict. I figured in-memory structures are best handled manually.

I also again made a lot of miscellaneous changes. Please excuse the clutter.😅

The tests should cover most of the ProteinStructureStore code. I am not sure if by "regression" you meant performance regression or just failing code. I saw BioStructures.jl has a suite of benchmarks. A similar setup could be used to automatically report performance regressions of pull requests.

[bicycle1885]

This looks good to me. By regression I meant avoiding functionality failures, not performance issues.