S3 class for estimate_infections

[sbfnk]

from the design doc:

• samples: get_samples() (or extract_samples()) built on the existing extract_parameter_samples(); this could also take a parameter name and look up the ID, or take a delay distribution name and look up the ID;
• summarised: summary()
• fit: get_stan_fit()
• args: not needed (only internally to identify parameter ids)?
• observations: not needed (only internally by summary() and get_samples()

[seabbs]

It looks like the direction of travel is get_ so makes sense to standarise around that vs extract I think