Adding the fit_model function

[gareth-j]

• Summary of changes (Bug fix, feature, docs update, ...)

This will add a fit_model function that in one line will create

• a mesh
• a formula
• fit the model
• return the fitted model object

• Please check if the PR fulfills these requirements

• Closes fit_model - generate formula #275
• Closes fit_model - meshing function #274
• Closes fit_model - fitting the model #276
• Tests added and passed
• All code checks passing - styler run over code
• Vignettes updated/added
• Added an entry in the latest CHANGELOG.md file if fixing a bug or adding a new feature
• Any new libraries added to DESCRIPTION

[XueqingYin]

Hi , I found several issues in fit_model.R and fit_model.Rmd.

for fit_model.R:

1. The purpose of having two inlabru::like() in this model is not clear to me. Maybe it's useful to explain the aim of each like(), and the problem that this model is solving?
2. Do you want to include the arguments (such as max.edge) that control mesh resolution when creating mesh? At present, they are missing in fm_mesh_2d_inla(), and I found max_edge_mesh is not used.
3. The object types for space_variables, spacetime_variables, and data need clear description. If space_variables and spacetime_variables are list objects as described now at the beginning of the script, then data[space_variables] will give an error "Error in .subset2(x, i, exact = exact) : invalid subscript type 'list'". To avoid this, space_variables and spacetime_variables should be a vector of spatial and spacetime variables, respectively.
4. The input for the argument locations should be a matrix of point locations, rather than a list of coordinates

for fit_model.Rmd:

1. process_coords argument does not exist in the fit_model() function
2. space_variables and spacetime_variables are missing in this rain example
3. If I remember it correctly, inlabru::bru requires the time index to start from 1, whereas it now starts from 5 in this example

[mnky9800n]

For the fit_model.rmd we actually stopped using that and it should be removed because we switched to using the test_fit_model.rmd

[mnky9800n]

Also, process_coords was replaced with locations to stay consistent

[mnky9800n]

We didn't put the arguments max.edge in the fmesher function because it is now supposed to estimate the best values for those. Since the goal was trying to produce a quick and dirty latent random field generator we figured this was inline with that idea.