make dynamic-ccall detection more correct

[vtjnash]

Relies on JuliaLang/julia#59165 for exact
accuracy, otherwise we're relying on hoping that eval on a random
expression is safe and correct (both hopes are quite false).

[codecov]

Codecov Report

❌ Patch coverage is 80.00000% with 3 lines in your changes missing coverage. Please review.
✅ Project coverage is 84.82%. Comparing base (376d172) to head (f10bd03).
⚠️ Report is 77 commits behind head on master.

Files with missing lines	Patch %	Lines
src/optimize.jl	80.00%	3 Missing ⚠️

Additional details and impacted files

@@            Coverage Diff             @@
##           master     #700      +/-   ##
==========================================
+ Coverage   81.03%   84.82%   +3.79%     
==========================================
  Files          12       12              
  Lines        2636     2728      +92     
==========================================
+ Hits         2136     2314     +178     
+ Misses        500      414      -86     

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.

[vtjnash]

merge?

[timholy]

A test that doesn't work currently would be lovely. JuliaInterpreter has s tendency to break each new Julia release, the tests are just about the only thing that prevent the whole thing from burning to the ground.

Honestly, now my preference would be to stop trying to interpret anything with a ccall, but keeping this working for now is also fine.

[vtjnash]

This is mainly about supporting JuliaLang/julia#59165, which removes the need for the unreliable hacks since they become syntactic instead of heuristic operations. Without this, it would start failing all tests once that is merged