chore: post Go 1.19 multi-error cleanup

[rvagg]

• replace some uses of multierr with fmt.Errorf's %w;%w
• fix up timeout error string inconsistencies: "timed out" vs "timeout"

as a result of checking the errors list with "Other" taking up a very large share, we have, in places 2 and 3:

 retrieval failed; retrieval timed out after 20s
 timeout after 20s

tbh I'm not actually sure why the former is now showing up more than the latter, I suspect it must have been something I did with the context juggling I've had to do

• fix: don't record failures on context cancellation #374
• fix: properly detect context cancellation error type #385

We're also getting this now:

 failed to load root node: failed to load root CID: context canceled

Which is fine, but I think it would be preferable to show this as a timeout instead (i.e. parent retrieval has a timeout and cancels context, the protocol retriever records it as a context cancel). It may be possible to use WithCancelCause to check for DeadlineExceeded post 1.19, but this'll need some investigation.

[codecov-commenter]

Codecov Report

Merging #453 (8643487) into main (2cf1121) will decrease coverage by 0.23%.
The diff coverage is 83.33%.

Additional details and impacted files

@@            Coverage Diff             @@
##             main     #453      +/-   ##
==========================================
- Coverage   75.73%   75.51%   -0.23%     
==========================================
  Files          87       87              
  Lines        6677     6657      -20     
==========================================
- Hits         5057     5027      -30     
- Misses       1335     1343       +8     
- Partials      285      287       +2     

Files	Coverage Δ
pkg/indexerlookup/candidatefinder.go	65.21% <100.00%> (-0.75%)	⬇️
pkg/retriever/bitswapretriever.go	95.83% <100.00%> (-0.14%)	⬇️
pkg/retriever/parallelpeerretriever.go	90.80% <100.00%> (-0.06%)	⬇️
pkg/retriever/graphsyncretriever.go	85.62% <66.66%> (-0.80%)	⬇️

... and 3 files with indirect coverage changes