guile: dynamic ffi (slow)

[antler5]

Hi! This pull isn't merge-worthy quite yet, and I don't know that I'll come back to it tomorrow, but I wanted to be sure to leave what I have in case anyone else has any insights.

I first implemented this as shown on the guile_concise branch, and then again with a dedicated extension and the extra boilerplate needed to compile it, in the naive hope that would speed it up, but I just don't know what I'm doing :p

Here are the results I get (with the tool-chains that I bothered to setup, am on GuixSD):

The results are elapsed time in milliseconds
============================================
luajit:
2283
2260
c:
3117
3093
cpp:
3092
3155
guile:
564329000 # ⎤ concise branch
464285000 # ⎦
354308000 # ⎤ extension branch
562453000 # ⎦
haskell:
3179
3141
node:
18180
17727
go:
76953
76586
julia:
4013
2239

Yeah, yikes, that can't be right. Maybe I'm measuring nanoseconds wrong? It went through a few drafts, but was always putting out a few figures more than I'd like., and really does take 10-15 seconds. I tried profiling it with statprof, but it's all too far down in plusone. I could try enabling some extra compiler optimizations (defaults to level 2 out of 9)?

What's missing:

• Why so slow???
• Readme updates
• Testing with upstream's configuration
• Guile auto-compiles .scm files on first-run, so it probably doesn't even need handled in the Tupfile for the concise branch. We could stick a shebang on guile/hello.scm, chmod +x, and simplify the call in run-all.sh to ./gulie/hello.scm, too, but I'm not cleaning up until I've gotten to the bottom the order-of-magnitude speed difference I'm seeing.