go.mod: update dependencies

[thaJeztah]

go.mod: github.com/pkg/errors v0.9.1

This version provides compatibility with Go 1.3+ native error wrapping

go.mod: github.com/google/go-cmp v0.5.5

Version v0.5.x provides various improvements to the reporter output;
https://github.com/google/go-cmp/releases

full diff: google/go-cmp@v0.4.0...v0.5.5

[thaJeztah]

Ah, looks like some changes are needed in a test (possibly different output format?)

=== RUN   TestPathField
    opt_test.go:249: assertion failed: 
        --- matches
        +++ expected
          []string{
          	"{opt.node}.Value.Value",
        + 	"{opt.node}.Children[0].Value.Value",
        + 	"{opt.node}.Children[1].Value.Value",
          	"{opt.node}.Children[2].Value.Value",
          	"{opt.node}.Children[2].Ref.Value.Value",
        - 	"{opt.node}.Children[1].Value.Value",
        - 	"{opt.node}.Children[0].Value.Value",
          	"{opt.node}.Children[0].Value.Value",
          	"{opt.node}.Children[1].Value.Value",
          	... // 2 identical elements
          }
        
--- FAIL: TestPathField (0.00s)

edit: breakage is in golang.org/x/tools; perhaps better to include in #218

[thaJeztah]

sigh.. so it was just random; issue looks to be in go-cmp

[thaJeztah]

Looks like go-cmp filter now may result in results to be returned in
random random order; this sorts the results before comparing, but I'm
not sure if order is important (in which case this could actually be
a bug)

[thaJeztah]

looks like CI is hitting concurrency limits 😞

[dnephin]

LGTM

The order shouldn't matter, because the filter will be applied to all matches. These filters have been difficult to test. It is not the first time that an upgrade has broken these tests.

[thaJeztah]

Thanks for confirming; I wasn't entirely sure, so thought I'd best call it out 👍 🤗