caliper_relevance_dynamic: Dynamic caliper width determination function
In ooelrich/oscbvar: Stuff for Bayesian vector autoregressions

This function dynamically choses the caliper width and calculates weights.

caliper_relevance_dynamic(
  atomic_df,
  sotw,
  start_agg = 401,
  calip_data,
  mvc = 1
)

`atomic_df`	Data frame of atomic predictions.
`sotw`	Data frame of smoothing variables. First column should be called "t", and should contain an unbroken sequence of integers.
`start_agg`	For which time point to start aggregating (what does this even mean??).
`calip_data`	A data table that contains two columns: time (t) and the optimal caliper width based on all predictions previous to this time (cw). That is, cw is the caliper width the decision maker should use at time t to pick the best caliper method based on historical data (from BERFORE t).
`mvc`	The minimum viable cluster parameter used by the caliper method.