gam_loglikelihood: Compute negative log-likelihood of gamma HMM parameters
In simonecollier/lizardHMM: Fitting Lizard Movement Data with HMMs

This function computes the negative log-likelihood that the given gamma HMM parameters could have generated the data being fit.

gam_loglikelihood(
  working_params,
  x,
  design,
  num_states,
  num_variables,
  num_subjects,
  num_covariates,
  state_dep_dist_pooled = FALSE
)

`working_params`	A vector of the working gamma parameters for the HMM.
`x`	The data to be fit with an HMM in the form of a 3D array. The first index (row) corresponds to time, the second (column) to the variable number, and the third (matrix number) to the subject number.
`design`	A list of design matrices for each subject with each row indicating the time and each column indicating the value of the covariate.
`num_states`	The number of states in the desired HMM.
`num_variables`	The number of variables in the data.
`num_subjects`	The number of subjects/trials that generated the data.
`num_covariates`	The number of covariates in the data that the transition probability matrix depends on.
`state_dep_dist_pooled`	A logical variable indiacting whether the state dependent distribution parameters `alpha` and `theta` should be treated as equal for all subjects.