sdreport comments

[kaskr]

Comments from @williamaeberhard :

• the obj$env$last.par shouldn't change after running sdreport on it, in particular it should not replace the estimates from the optimizer (the values we really should keep) by the perturbed ones coming from the finite-differences approx of optimHess.
• sdreport prints some things that may be misleading, especially when supplying the Hessian with hessian.fixed.