Implement analytical solutions in R and stan

[seabbs]

We need

• R framework to handle cases where we do and don't have analytically solutions + a switch to override and use numerical (for testing/user confidence)
• Stan framework to handle cases where we do and don't have analytical solutions. This may not need an override switch or we may at least not need to propagate it
• Implement log version of gamma with uniform
• Implement log version of lognormal with uniform
• Make sure all current tests are passing and benchmarks are faster as expected
• Either add explicit new tests for numerical coverage or make all current tests use the override switch and write new tests to test analytical vs numeric (I think I favour the latter)
• New benchmarks so we cover both analytical and numeric models.

[seabbs]

Work on R side analytical infra is going in here: https://github.com/epinowcast/primarycensoreddist/tree/analytical-infra

using a s3 based approach that dispatches on dist combinations with an override

[seabbs]

Pulling Weibull it its own issue