fix CSV unmarshaling error with custom unmarshaler with pretty errors

[grokspawn]

Description of the change:

Solves #1039 for bundle property rendering cases.

For bundles with invalid CSVs we would see unhelpful messages like

opm render registry-proxy.engineering.redhat.com/rh-osbs/kmm-kernel-module-management-operator-bundle@sha256:a55eb76e830293c579c4935bf74b9ab5c0ab9f5bcd9a324ae23a797976d86695
2022/11/10 15:01:34 render reference "registry-proxy.engineering.redhat.com/rh-osbs/kmm-kernel-module-management-operator-bundle@sha256:a55eb76e830293c579c4935bf74b9ab5c0ab9f5bcd9a324ae23a797976d86695": unrecognized type: string

This change allows us to catch/format unmarshaling errors of the CSV to provide messages like the following, where there is a marker showing the error as well as more information about the nature of the error (shortened for legibility):

./bin/opm render registry-proxy.engineering.redhat.com/rh-osbs/kmm-kernel-module-management-operator-bundle@sha256:a55eb76e830293c579c4935bf74b9ab5c0ab9f5bcd9a324ae23a797976d86695
2023/08/21 16:21:45 render reference "registry-proxy.engineering.redhat.com/rh-osbs/kmm-kernel-module-management-operator-bundle@sha256:a55eb76e830293c579c4935bf74b9ab5c0ab9f5bcd9a324ae23a797976d86695": json: cannot unmarshal bool into Go struct field ObjectMeta.metadata.annotations of type string at offset 344 (indicated by <==)
{
    "apiVersion": "operators.coreos.com/v1alpha1",
    "kind": "ClusterServiceVersion",
    "metadata": {
        "annotations": {
            "alm-examples": "[\n  {\n    \"apiVersion\": \"kmm.sigs.k8s.io/v1beta1\",\n    \"kind\": \"Module\",\n    \"metadata\": {\n      \"name\": \"module-sample\"\n    },\n    \"spec\": null\n  }\n]",
            "capabilities": "Basic Install",
            "certified": true <== ,
            "containerImage": "registry.redhat.io/kmm/kernel-module-management-operator-rhel8@sha256:fddffaaac1b81ee29bf94af94373c15a7def1b14e34e92ce171aa791022f9ffd",
            "olm.skipRange": "\u003e=0.0.0 \u003c1.0.0",
            "operators.operatorframework.io/builder": "operator-sdk-v1.17.0+git",
            "operators.operatorframework.io/project_layout": "go.kubebuilder.io/v3",
            "repository": "https://github.com/rh-ecosystem-edge/kernel-module-management",
            "support": "Red Hat"
        },
        "name": "kmm-operator.v1.0.0",
        "namespace": "placeholder"
    },

 ---- SNIP ----

Motivation for the change:

Reviewer Checklist

• Implementation matches the proposed design, or proposal is updated to match implementation
• Sufficient unit test coverage
• Sufficient end-to-end test coverage
• Docs updated or added to /docs
• Commit messages sensible and descriptive

[grokspawn]

/hold until I get the utests sorted

[codecov]

Codecov Report

Merging #1138 (c93d7b8) into master (dba1e10) will increase coverage by 0.01%.
Report is 1 commits behind head on master.
The diff coverage is 58.82%.

❗ Current head c93d7b8 differs from pull request most recent head b7b447f. Consider uploading reports for the commit b7b447f to get more accurate results

@@            Coverage Diff             @@
##           master    #1138      +/-   ##
==========================================
+ Coverage   53.79%   53.81%   +0.01%     
==========================================
  Files         108      108              
  Lines       10166    10195      +29     
==========================================
+ Hits         5469     5486      +17     
- Misses       3736     3741       +5     
- Partials      961      968       +7     

Files Changed	Coverage Δ
pkg/registry/csv.go	69.23% <53.33%> (-3.32%)	⬇️
alpha/declcfg/declcfg.go	80.48% <100.00%> (ø)
pkg/prettyunmarshaler/prettyunmarshaler.go	89.74% <100.00%> (ø)

... and 1 file with indirect coverage changes

[grokspawn]

I sorted out the sequence, and maybe the multiple 'helper' functions aren't necessary, but it aided my investigation. It turns out that because there is only one named field in the struct, and it is a json.RawMessage we have to be careful to extract it by name.

Now only a single utest is failing: TestExportForBundleWithBadImage, because it attempts a registry pull of image foo and it never comes back from syscall.

I'm investigating, but here's the trace in case something jumps out to someone.

goroutine 69 [syscall, 10 minutes]:
syscall.syscall6(0xc000723728?, 0x104bd5e94?, 0x1081818e0?, 0x10986cf40?, 0xc0007237a8?, 0x104bd60fc?, 0x104b8bfd0?)
	/Users/jordan/.gvm/gos/go1.20.7/src/runtime/sys_darwin.go:45 +0x68
syscall.wait4(0x108149640?, 0xc0007237a8, 0x0?, 0xc000726000)
	/Users/jordan/.gvm/gos/go1.20.7/src/syscall/zsyscall_darwin_arm64.go:43 +0x64
syscall.Wait4(0xc0002ba318?, 0xc000723808, 0xc000121b30?, 0xc000294660?)
	/Users/jordan/.gvm/gos/go1.20.7/src/syscall/syscall_bsd.go:144 +0x48
os.(*Process).wait(0xc000294660)
	/Users/jordan/.gvm/gos/go1.20.7/src/os/exec_unix.go:43 +0xbc
os.(*Process).Wait(...)
	/Users/jordan/.gvm/gos/go1.20.7/src/os/exec.go:132
os/exec.(*Cmd).Wait(0xc000121a20)
	/Users/jordan/.gvm/gos/go1.20.7/src/os/exec/exec.go:890 +0x74
os/exec.(*Cmd).Run(0xc00011e930?)
	/Users/jordan/.gvm/gos/go1.20.7/src/os/exec/exec.go:590 +0x58
os/exec.(*Cmd).CombinedOutput(0xc000121a20)
	/Users/jordan/.gvm/gos/go1.20.7/src/os/exec/exec.go:1005 +0x218
github.com/operator-framework/operator-registry/pkg/containertools.(*ContainerCommandRunner).Pull(0xc00000d560, {0x1069093d1, 0x3})
	/Users/jordan/devel/operator-registry/pkg/containertools/runner.go:90 +0x318
github.com/operator-framework/operator-registry/pkg/image/execregistry.(*Registry).Pull(0xc00000d590, {0x3?, 0xc00070dd10?}, {0x1071d3b68, 0xc0003782b0})
	/Users/jordan/devel/operator-registry/pkg/image/execregistry/registry.go:38 +0x78
github.com/operator-framework/operator-registry/pkg/lib/bundle.(*BundleExporter).Export(0xc000723e60, 0x1, 0x0)
	/Users/jordan/devel/operator-registry/pkg/lib/bundle/exporter.go:67 +0x4c0
github.com/operator-framework/operator-registry/pkg/lib/bundle.TestExportForBundleWithBadImage(0x0?)
	/Users/jordan/devel/operator-registry/pkg/lib/bundle/exporter_test.go:12 +0xb0
testing.tRunner(0xc0001a0680, 0x1071c7878)
	/Users/jordan/.gvm/gos/go1.20.7/src/testing/testing.go:1576 +0x18c
created by testing.(*T).Run
	/Users/jordan/.gvm/gos/go1.20.7/src/testing/testing.go:1629 +0x5e8

[grokspawn]

/unhold

Now only a single utest is failing: TestExportForBundleWithBadImage, because it attempts a registry pull of image foo and it never comes back from syscall.

Evidently this issue only exists on mac(??) and it runs cleanly on linux. Going to get this into review while I continue to investigate.

[grokspawn]

/unapprove
I'd really like to get two sets of eyeballs on this, so I'm removing my (default) approval.

[grokspawn]

/remove-approve

[grokspawn]

Now only a single utest is failing: TestExportForBundleWithBadImage, because it attempts a registry pull of image foo and it never comes back from syscall.

Evidently this issue only exists on mac(??) and it runs cleanly on linux. Going to get this into review while I continue to investigate.

Turns out to be a bad docker install on my mac. :(
Resolved, and all tests pass now.

[awgreene]

/approve

[openshift-ci]

[APPROVALNOTIFIER] This PR is APPROVED

This pull-request has been approved by: awgreene, everettraven

The full list of commands accepted by this bot can be found here.

The pull request process is described here

Needs approval from an approver in each of these files:

• OWNERS [awgreene]

Approvers can indicate their approval by writing /approve in a comment
Approvers can cancel approval by writing /approve cancel in a comment

[awgreene]

Ah @everettraven already approved
/lgtm