Improve inferrability
There is no way to make the S keyword argument of LinearOperator{T}
inferrable.
This moves S to the second position, so anyone who wishes to write
inferrable code can use LinearOperator{T, Vector{T}}(args...).
This appears to be non-breaking. A search of the JuliaSmoothOptimizers org for code with LinearOperator{ reveals only one source-code usage with more than one parameter: https://github.com/JuliaSmoothOptimizers/RipQP.jl/blob/6d3c074e72b4e94846a165bed5cc2b78bfdb3623/src/iterations/preconditioners/LDL.jl#L48
While the names there will be misleading, there appears to be no functional consequence.
Bump
Any chance someone can take a look at this? Or at least approve the workflow so that the tests can run? (EDIT: now I'm a bit confused about whether this repo actually uses the CI workflow.)