Minres Fail to solve subproblem and cuases R2N solver to Max_timed

Convergence tolerance is more demanding when using warm-start

Add a cross-reference for KrylovConstructor in the Krylov workspaces

Unsupported keyword argument "memory" for `GMRES` and `DQGMRES`

Another lnlq NaN case

Worse convergence for `block_minres` than `minres`

Wanted: type indicating efficient M and M⁻¹

Documentation confusion between preconditioners and the problem being solved

New PR template has some typos

Check the unit tests of TriLQR (and test_warm_start.jl)