Add jumps to particle state

[THargreaves]

A modification of the existing Levy SSM example to include the jumps (intensity and times) in the state.

This is important for performing inference on the jump times/location. This is largely in preparation for the future when the jump times/locations are sampled using a more intricate method than forward simulation (Yaman has a simple SMCMC approach, but there are probably more nuanced approaches).

I've taken quite a hacky approach, but I'm struggling to see another way. Specifically, AdvancedPS expects transition to return a distribution rather than a new state. This meant that I needed to define a TransitionDistribution struct with a rand method that runs the sampling.

Another approach is creating a Dirac point distribution at the new state but that feels like an unstable approach to me.

Changing AdvancedPS to expect a new sample to be returned from transition also doesn't quite work. For PG this is fine but for PGAS we need to calculate the log prob of the transition distribution too.

[codecov]

Codecov Report

All modified and coverable lines are covered by tests ✅

Comparison is base (2902fa6) 96.09% compared to head (76ff740) 95.86%.

Additional details and impacted files

@@             Coverage Diff              @@
##           levy-ssm      #96      +/-   ##
============================================
- Coverage     96.09%   95.86%   -0.24%     
============================================
  Files             8        8              
  Lines           410      411       +1     
============================================
  Hits            394      394              
- Misses           16       17       +1     

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.