Add support for constructing target distribution from formula

[matt-graham]

Adds support for specifying target distribution by specifying a formula for log density using R ~ formula syntax (only right-hand side of formula is used). A function to compute log density and its gradient is then constructed using the stats::deriv function along with a function to trace variables from chain state.

As well as adding a utility function for this conversion, this PR also changes the behaviour of sample_chain to allow accepting formulas and bridgestan Stan models directly as the target_distribution argument, with these then converted internally to a list of functions. As part of this the separate trace_function optional argument has been removed in favour of assuming this is specified as part of the target_distribution list (or constructed from the formula / Stan model).

[codecov]

Codecov Report

All modified and coverable lines are covered by tests ✅

Project coverage is 100.00%. Comparing base (95acd2d) to head (6efa4f1).
Report is 1 commits behind head on main.

Additional details and impacted files

@@            Coverage Diff            @@
##              main       #66   +/-   ##
=========================================
  Coverage   100.00%   100.00%           
=========================================
  Files           11        11           
  Lines          525       564   +39     
=========================================
+ Hits           525       564   +39     

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