Allow specifying sigma in LinearMixedModel

[palday]

This was done to support meta-analytic type models. There might be other cases where we can use a fixed residual variation to do tricks with observation-level random effects, but I'm not sure.

I've also done some formatting tweaks that take advantage of kwarg syntax in Julia 1.5+ (since we now require Julia 1.6).

I'll tackle #549 after this is merged (or rejected) as some of the changes would create a merge conflict.

If we one day allow specifying constraints on the betas, I would suggest doing this as a new type ConstrainedLinearMixedModel and then define a separate objective method for that so that we don't keep adding conditionals to objective. (And reallistically, we should should also define separate setβ!, setθ! and setβθ! methods.)

[codecov]

Codecov Report

Merging #551 (3a26237) into main (5fe9112) will increase coverage by 0.09%.
The diff coverage is 100.00%.

❗ Current head 3a26237 differs from pull request most recent head 83bc987. Consider uploading reports for the commit 83bc987 to get more accurate results

@@            Coverage Diff             @@
##             main     #551      +/-   ##
==========================================
+ Coverage   96.16%   96.25%   +0.09%     
==========================================
  Files          27       27              
  Lines        2425     2431       +6     
==========================================
+ Hits         2332     2340       +8     
+ Misses         93       91       -2     

Impacted Files	Coverage Δ
src/optsummary.jl	97.67% <ø> (ø)
src/linearmixedmodel.jl	98.43% <100.00%> (+0.01%)	⬆️
src/linalg/cholUnblocked.jl	100.00% <0.00%> (+9.09%)	⬆️

---

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update 5fe9112...83bc987. Read the comment docs.