Fix various issues in initializing nimOptim, plus use of Laplace updateSettings

[paciorek]

See NCT 571 for a full description of the bug and the necessary changes.

[paciorek]

Note that in Laplace we should only set fnscale to -1 if it is set to the default value of 1, so as to allow users to set it to something different (e.g., -10). I've done that in nimbleQuad, but not here.

[perrydv]

@paciorek Thanks for catching the needs for as.numeric and working_maxit.

My idea with allowing replacement of the optimControlList was that it is an advanced thing to do, and so if a user does it, they must attend to all elements of the optimControlList. I did not want to put in the effort to somehow allow changing just individual elements, and I suspect it would get clunky. So it could be little discordant to stick in fnscale=-1, i.e. to force one element to one value that we think must be right. While I can't imagine a case where it should be positive, could they experiment with fnscale = -0.5 or -10? Instead of forcing it to -1, would we want to check if it's non-negative and throw an error or warning if so? Of course, it is always possible to intervene in control settings by providing one's own optimizer (which could call a standard optimizer but change control values or other settings). -1 is likely safe and helpful in many or most cases, so we could stick with it.

[paciorek]

@perrydv sorry to be confusing about the fnscale business. Here in nimble, I just stick in -1, which as you say is too prescriptive. But in nimbleQuad I only stick in -1 if the value is 1. And the code in nimble will disappear when I delete the Laplace code for the 1.4.0 release. So I think we are now on the same page.