vendor: github.com/urfave/cli v1.22.9 and fix "verify-vendor" script

[thaJeztah]

fix verify-vendor if go.mod does not contain replace rules

Before this patch, if the go.mod did not contain a replace rule, jq would fail
(Cannot iterate over null), and Bash would produce an error when trying to add
an empty $key to map_requires:

make verify-vendor
+ verify-vendor
  jq: error (at <stdin>:581): Cannot iterate over null (null)
  /go/src/github.com/containerd/containerd//script/verify-go-modules.sh: line 44: map_replaces_1[$key]: bad array subscript
  make: *** [Makefile:435: verify-vendor] Error 1

vendor: github.com/urfave/cli v1.22.9

The regression in v1.22.2 has been resolved, so we can drop the
replace rule and use the latest v1.22.x version.

full diff: urfave/cli@v1.22.1...v1.22.9

[thaJeztah]

I'll work on a follow-up to this one to see if we can make use of the new urfave_cli_no_docs build-tag as well (see urfave/cli#1383)

[thaJeztah]

Hmm... what's this error?

+ verify-vendor
go: downloading github.com/Microsoft/hcsshim/test v0.0.0-20210408205431-da33ecd607e1
jq: error (at <stdin>:581): Cannot iterate over null (null)
/home/runner/work/containerd/containerd/src/github.com/containerd/containerd//script/verify-go-modules.sh: line 44: map_replaces_1[$key]: bad array subscript
Makefile:422: recipe for target 'verify-vendor' failed
make: *** [verify-vendor] Error 1

[thaJeztah]

Pushed a commit to fix the vendor check

[thaJeztah]

Failures should probably not be related (as #6998 went green); let me close/reopen to trigger CI

[thaJeztah]

Hmm.. that didn't seem to trigger the cirrus jobs (which were the ones failing); perhaps it's just slow to report, so let's see

[estesp]

LGTM 🎉 (nice to see the replace rules disappearing!)