package extensions - weak deps

[t-bltg]

Fix #4589.

Note that we don't replace the @require statements for backends loading since their treatment is somewhat "special":

Plots.jl/src/backends.jl

Line 100 in ef50445

backend_name() === $be || @require $pkg = $(_plots_deps["$pkg"]) begin

I'm unsure this logic can be moved to package extensions (neither if it's worth it).

NOTE: the formatter check failure is investigated in tkf/JuliaProjectFormatter.jl#7.

[codecov]

Codecov Report

Base: 90.48% // Head: 90.43% // Decreases project coverage by -0.06% ⚠️

Coverage data is based on head (65af264) compared to base (74d8d53).
Patch coverage: 100.00% of modified lines in pull request are covered.

Additional details and impacted files

@@            Coverage Diff             @@
##           master    #4649      +/-   ##
==========================================
- Coverage   90.48%   90.43%   -0.06%     
==========================================
  Files          41       40       -1     
  Lines        8801     8700     -101     
==========================================
- Hits         7964     7868      -96     
+ Misses        837      832       -5     

Impacted Files	Coverage Δ
src/Plots.jl	88.88% <ø> (ø)
src/output.jl	97.32% <ø> (+1.63%)	⬆️
src/components.jl	90.62% <100.00%> (ø)
src/init.jl	100.00% <100.00%> (ø)
src/utils.jl	92.75% <100.00%> (+0.09%)	⬆️
src/backends/pythonplot.jl	87.67% <0.00%> (-0.63%)	⬇️
ext/UnitfulExt.jl
src/layouts.jl	95.75% <0.00%> (+0.28%)	⬆️
RecipesPipeline/src/group.jl	94.59% <0.00%> (+4.05%)	⬆️

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

☔ View full report at Codecov.
📢 Do you have feedback about the report comment? Let us know in this issue.

[t-bltg]

I fail to see:

1. why this PR increases (slightly) invalidations - not blocking;
2. most importantly, why CI segfaults on PythonPlot related stuff on linux - ubuntu after testing ends, while a dummy PR opened at the same moment passes CI (dummy pr #4650) - this is blocking.

I cleaned up the actions caches, but it didn't seem to help 🤷 .
It also does not segfault locally.

I guess another look at this (maybe @mkitti, or @BeastyBlacksmith) is necessary.

[t-bltg]

Setting the ENV variable MPLBACKEND: "agg" seems to solve the segfaults in ci.
And the invalidations resolved themselves automagically, love that !

Review comments welcome now.

[t-bltg]

I'll probably merge this PR in the next few days, unless new comments are made.