Consider -Wconversion warnings

Enabling -Wconversion (particularly with clang++ as on the M1mac machine at CRAN, see its README) exposes a number of warnings. The Oxford machine uses clang++-21, on Ubuntu 26.04 we also have clang++-22 which Vienna uses (without the same options). Per the README this uses

CFLAGS="-falign-functions=8 -g -O2 -Wall -pedantic -Wconversion -Wno-sign-conversion -Wstrict-prototypes"
C17FLAGS="-falign-functions=8 -g -O2 -Wall -pedantic -Wconversion -Wno-sign-conversion -Wno-strict-prototypes"
# ...
CXXFLAGS="-g -O2 -Wall -pedantic -Wconversion -Wno-sign-conversion"

Note that this turns 'sign-conversion' warnings off. Those account for a ton. I found first set of changes we could make by not doing this i.e. by running with only -Wconversion. That may be too radical, but it could still hide some overflows.

Also, our own compilation is 'clean' under -Wconversion -Wno-sign-conversion however other package may see

In file included from /Users/ripley/R/Library/Rcpp/include/Rcpp/sugar/functions/functions.h:63:
/Users/ripley/R/Library/Rcpp/include/Rcpp/sugar/functions/mean.h:38:14: warning: implicit conversion from 'R_xlen_t' (aka 'long') to 'long double' may lose precision [-Wimplicit-int-float-conversion]
   38 |         s /= n;
      |           ~~ ^
/Users/ripley/R/Library/Rcpp/include/Rcpp/sugar/functions/mean.h:44:20: warning: implicit conversion from 'R_xlen_t' (aka 'long') to 'long double' may lose precision [-Wimplicit-int-float-conversion]
   44 |             s += t/n;
      |                   ~^

Start by reproducing the reported build with clang++ and the -Wconversion flags from the M1mac README. Inspect the warnings, including the example in Rcpp/sugar/functions/mean.h, and determine the agreed scope; done means the project has a documented decision and the selected warnings are resolved or intentionally excluded.