Optimize visiting code

[krlmlr]

For #558 and #759.

This is a tad faster and enables cleaner stack traces, in particular recursive calls now follow each other directly, without an interleaving map(). We now can profile, flatten recursive calls, and get a clean picture where time is actually spent.

[codecov-commenter]

Codecov Report

Merging #979 (7031c22) into main (2fa7369) will decrease coverage by 0.02%.
The diff coverage is 90.69%.

@@            Coverage Diff             @@
##             main     #979      +/-   ##
==========================================
- Coverage   90.12%   90.10%   -0.03%     
==========================================
  Files          47       47              
  Lines        2664     2698      +34     
==========================================
+ Hits         2401     2431      +30     
- Misses        263      267       +4     

Impacted Files	Coverage Δ
R/initialize.R	96.96% <ø> (ø)
R/relevel.R	47.82% <50.00%> (-0.07%)	⬇️
R/visit.R	94.84% <92.30%> (-5.16%)	⬇️
R/nested-to-tree.R	96.29% <100.00%> (ø)
R/transform-block.R	100.00% <100.00%> (ø)

Help us with your feedback. Take ten seconds to tell us how you rate us. Have a feature suggestion? Share it here.

[github-actions]

This is how benchmark results would change (along with a 95% confidence interval in relative change) if d8a7301 is merged into main:

•   :rocket:cache_applying: 26.8ms -> 24ms [-11.02%, -9.94%]
•   :rocket:cache_recording: 1.11s -> 1.09s [-2.33%, -1.26%]
•   :rocket:without_cache: 2.96s -> 2.88s [-2.85%, -2.09%]

Further explanation regarding interpretation and methodology can be found in the documentation.

[lorenzwalthert]

Nice, speed benchmarks support your claim. I am sold 🥳