`get_variance()` Triggers Model Recompilation and Returns `NULL` for `brms` Mixed Models in v0.20.3

[matschmitz]

When using get_variance() from insight v0.20.3, the model unexpectedly recompiles, and the function returns NULL. This behavior differs from v0.20.1, where it correctly returns the variance components without recompiling the model.

library(brms)
library(insight)

mdl <- brm(mpg ~ hp + (1 | cyl), data = mtcars)

get_variance(mdl)  # Model recompiles, returns NULL with v0.20.3
# Model recompiles
# NULL

get_variance(mdl)  # Returns expected variance components with v0.20.1
# $var.fixed
# [1] 4.90803
# 
# $var.random
# [1] 22.64787
# 
# $var.residual
# [1] 10.82719
# 
# $var.distribution
# [1] 10.82719
# 
# $var.dispersion
# [1] 0
# 
# $var.intercept
# cyl 
# 22.64787 

However, when using a model fitted with lme4::lmer, get_variance() returns the expected output, even in v0.20.3:

library(lme4)
library(insight)

mdl <- lmer(mpg ~ hp + (1 | cyl), data = mtcars)

get_variance(mdl)  # Returns expected variance components with v0.20.1
# $var.fixed
# [1] 4.363142
# 
# $var.random
# [1] 16.18444
# 
# $var.residual
# [1] 9.917004
# 
# $var.distribution
# [1] 9.917004
# 
# $var.dispersion
# [1] 0
# 
# $var.intercept
# cyl 
# 16.18444 

### Tasks