Linking undetected/possible undeclared dependencies

Reposting this as an issue. Don't have time to work this right now, but if anyone else is so motivated, here's my assessment:

> I think this recipe needs more scrutiny. The link analysis doesn't find that `pcre2` is actually linked to in the final library. Moreover, there are many more library flags than this recipe is declaring:
>
> ```bash
> ## [NB: line breaks added for readability]
> x86_64-apple-darwin13.4.0-clang -o libjri.jnilib Rengine.o jri.o \
> Rcallbacks.o Rinit.o globals.o rjava.o -Wl,-dead_strip_dylibs \
> -Wl,-pie -Wl,-headerpad_max_install_names -Wl,-dead_strip_dylibs \
> -Wl,-rpath,$PREFIX/lib -L$PREFIX/lib -Wl,-pie -Wl,-headerpad_max_install_names \
> -Wl,-dead_strip_dylibs -Wl,-rpath,$PREFIX/lib -L$PREFIX/lib -L$PREFIX/lib/R/lib \
> -dynamiclib -dynamiclib -Wl,-headerpad_max_install_names \
> -undefined dynamic_lookup -single_module -multiply_defined suppress \
> -L$PREFIX/lib/jvm/lib/server -ljvm -fopenmp -Wl,-dead_strip_dylibs -Wl,-pie \
> -Wl,-headerpad_max_install_names -Wl,-dead_strip_dylibs -Wl,-rpath,$PREFIX/lib \
> -L$PREFIX/lib -L$PREFIX/lib/R/lib -lR -L$PREFIX/lib -lpcre2-8 -llzma -lbz2 -lz -ldl \
> -lm -liconv -licuuc -licui18n
> ```
>
> I suspect either the linking analysis is not looking at `libjri.jnilib` or `pcre2` really isn't needed. In the former case, we probably should be declaring several more dependencies; in the latter, this migration is moot.
>
> Either way, I'd like to see this answered before merging.

_Originally posted by @mfansler in https://github.com/conda-forge/r-rjava-feedstock/pull/38#pullrequestreview-1738747654_