Silence sign conversion warnings from NClist functions

[ZedThree]

Fixes warnings about sign conversion from uses of nclistget/nclistlength and other NClist functions.

In a lot of the places that these functions are used, the index is an int instead of size_t, mostly in loops. In this PR, I've fixed these local variables to be size_t -- an alternative would be to change the API for NClist to use int, but unlike #2781 this still leaves >100 warnings, so there would still be lots of places that need fixing.

I've opted for changing the local variables to size_t rather than just casting the argument to nclistget, as this tended to fix multiple warnings in one go. There are a few places where casting is the better option, mostly where there's some interaction with a netCDF int ID.

There was one place (libdap2/getvara.c) where I also changed the function signature, but this was a local static function and I could verify all places where it is used.

I've left alone any place where there's some potentially unsafe arithmetic done with the list index, e.g. index - 1 or index--, mostly in loops iterating backwards over the list. I just wanted to go for the easy, low hanging fruit here.

This silences >400 warnings.

[DennisHeimbigner]

I think there is an error in this PR somewhere. When run thru github actions,
several tests are timing out.

[ZedThree]

I was a little rushed and missed a couple of places where it wasn't safe to just switch to size_t because e.g. some arithmetic was done with the result of nclistlength -- now fixed.

There's a few places where it would be safe to switch to size_t but then return/break if nclistlength(list) == 0. I've not made those changes here to try to keep this large PR simpler.

[ZedThree]

A much simpler alternative could be to change size_t to int in nclist.h/c completely. This changes ~20 lines and would silence ~250 warnings, although it might be then necessary to insert some >= 0 checks in places.

[WardF]

Testing something locally; I expect it to pass, but better safe than sorry. Thanks, once this passes I'll get it merged in!

[DennisHeimbigner]

might be then necessary to insert some >= 0 checks in places.

Definitely true in any function in nclist that is checking an incoming index into the list.