Replace the slow slotname Dict with a counter ("age")-based mechanism

[timholy]

Old benchmarks:

          "recursive other 1_000" => Trial(7.510 ms)
          "recursive self 1_000" => Trial(2.450 ms)
          "long function 5_000" => Trial(33.257 ms)
          "throw long 1_000" => Trial(14.002 ms)
          "tight loop 10_000" => Trial(213.216 ms)

vs this branch:

         "recursive other 1_000" => Trial(7.502 ms)
          "recursive self 1_000" => Trial(1.996 ms)
          "long function 5_000" => Trial(30.500 ms)
          "throw long 1_000" => Trial(13.406 ms)
          "tight loop 10_000" => Trial(174.004 ms)

Take with a grain of salt, there is a fair amount of variance. But the last one is consistently cut to about 80% of the original.

[codecov-io]

Codecov Report

Merging #307 into master will increase coverage by 0.3%.
The diff coverage is 100%.

@@            Coverage Diff            @@
##           master     #307     +/-   ##
=========================================
+ Coverage   88.17%   88.47%   +0.3%     
=========================================
  Files          11       11             
  Lines        1742     1753     +11     
=========================================
+ Hits         1536     1551     +15     
+ Misses        206      202      -4

Impacted Files	Coverage Δ
src/construct.jl	92.18% <100%> (-0.22%)	⬇️
src/types.jl	72.54% <100%> (+1.12%)	⬆️
src/breakpoints.jl	92.64% <100%> (+3.67%)	⬆️
src/interpret.jl	86.31% <100%> (ø)	⬆️
src/utils.jl	85.57% <100%> (+0.35%)	⬆️

---

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update f9ebc94...9bb6176. Read the comment docs.

[KristofferC]

Could potentially make this a concrete Int64 if we worry at all about overflow (on 32-bit)

[timholy]

Does the approval mean you know it doesn't break Debugger.jl? How about Juno?

[KristofferC]

It means the PR on its own is good but it is indeed breaking Debugger. Will fix that up tomorrow.

[KristofferC]

JuliaDebug/Debugger.jl#198 should fix the Debugger.jl side.

[timholy]

Let's wait for @pfitzseb to get a chance to look it over, in case it has more negatives than we might imagine.

[KristofferC]

JunoLab/Atom.jl#148 should fix the Juno side.

[KristofferC]

I think we should just put eval_code into this package.

[pfitzseb]

Agreed, it feels bad to have to keep both copies in sync...

[KristofferC]

Ok, with eval_code in this repo and updated in this PR I will merge this.