Add serialisation as an optional feature

[alex-mckenna]

This commit introduces an optional feature for serialising and
deserialising CPUID information using serde. Although this may seem
like an unnecessary addition, it can be useful when working with
distributed systems where some jobs require a minimal set of ISA
extensions. By allowing serialisation of the types in the library,
information about computers can be easily shared (without the need for
orphan instances or unnecessary newtype definitions).

[alex-mckenna]

Please let me know if there's anything else you require, such as bumping the version number etc.

[gz]

Thanks, seems like a very useful thing to have! Is it possible to add tests for it to make sure we don't break the feature in the future? It would be great if we can run the existing tests with the deserialized/serialized data.

I mostly want to avoid a case in the future where a new struct is introduced and we miss to add the derives.

[alex-mckenna]

No problem, I'll try and get some tests written in the next few days

[gz]

I'm going to merge this already since I'm currently writing a simple cpuid tool for the library and I will try and integrate the serialization there as well. In case you get a chance for the tests at some point feel free to open another pull request.

[gz]

Thanks for the patch!