Add 'replay' context generation method

[michaeltremeer]

This PR adds a second method for generating messages, so that users can load a list of messages payloads from file and use these for testing (which a number of customers are looking to use for real-world testing).

This PR:

• Adds a new context-generation-method arg to allow the user to select either generation or replay methods, along with a replay-path arg for users to define the path to the JSON file.
• Adds a messagegeneration.py module that contains an abstract base class for any new MessagesGenerators, refactors the existing _generate_messages function into a new class, and adds another class for the replay context generation method.
• Updates to the README

[technicianted]

The PR is about replay but then we're just using the dataset randomly which would not give the same usage patterns as the original dataset. For example, if I provide a dataset that has cyclical patterns in them that represent my shape over 4 hours for example, the results are not going to be what's expected.

Please correct me if I'm misunderstanding the use-case here.

I understand that this may add another layer of complexity: what about timing of the dataset? Originally I had thought that this was going to be your idea: in replay mode we don't provide RPM. Instead, we get timing from the dataset itself and we replay it the same way.

One proposal is to keep this as-is but name it something different and use the dataset sequentially rather than in random. Perhaps call it file. Then we can work on a completely different subcommand that would implement a true dataset replay.

What do you think?