Don't propagate widening axis limits for pipeline recipes

[t-bltg]

The bug described in #3451 is triggered by https://github.com/JuliaPlots/RecipesPipeline.jl/blob/d11dcb642340c4bbe4068130d09597b54c079bcd/src/user_recipe.jl#L249.

Applying a function without x values should only be applied on the initial x data, not on the widened axis data (causing log to fails with NaNs on negative values).

Fix #3451.

Mwe

using Plots; gr()

main() = begin
  plot([.1, 5], [5, .1], yscale=:log10)
  plot!(identity)
  png("bug")
end

main()

Output

[codecov]

Codecov Report

Merging #3619 (e620730) into master (0122d70) will decrease coverage by 1.72%.
The diff coverage is 14.48%.

❗ Current head e620730 differs from pull request most recent head 909ad33. Consider uploading reports for the commit 909ad33 to get more accurate results

@@            Coverage Diff             @@
##           master    #3619      +/-   ##
==========================================
- Coverage   65.15%   63.42%   -1.73%     
==========================================
  Files          27       28       +1     
  Lines        6672     6841     +169     
==========================================
- Hits         4347     4339       -8     
- Misses       2325     2502     +177     

Impacted Files	Coverage Δ
src/backends.jl	61.06% <0.00%> (-2.25%)	⬇️
src/backends/gaston.jl	0.00% <0.00%> (ø)
src/backends/gr.jl	89.12% <100.00%> (+0.11%)	⬆️
src/init.jl	100.00% <100.00%> (ø)
src/layouts.jl	67.70% <100.00%> (-0.61%)	⬇️
src/pipeline.jl	89.63% <100.00%> (+0.51%)	⬆️
src/plot.jl	67.71% <100.00%> (+0.51%)	⬆️
src/axes.jl	83.56% <0.00%> (-0.47%)	⬇️
src/examples.jl	61.76% <0.00%> (ø)
... and 4 more

---

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 c9f7c6b...909ad33. Read the comment docs.