Implement recursive SanitizedCopy() for KongState, refactor SetCredential

[mflendrich]

Please review commit by commit and merge without squashing.
Inspired by #991.

This PR:

• defines Go types for credential types in KIC (as wrappers for go-kong types allowing for implementation of KIC business logic in those wrappers) consistently with wrappers for other go-kong types (Consumer, Service, Route, etc.)
• implements a recursive Kongstate.SanitizedCopy() produces a copy of KongState containing only a subset of KongState fields,
• refactors SetCredential by breaking its type-specific logic out to individual type creator functions NewXXX()

[codecov]

Codecov Report

Merging #1021 (6c28445) into next (d2442de) will increase coverage by 1.55%.
The diff coverage is 85.44%.

@@            Coverage Diff             @@
##             next    #1021      +/-   ##
==========================================
+ Coverage   49.78%   51.33%   +1.55%     
==========================================
  Files          32       33       +1     
  Lines        3224     3327     +103     
==========================================
+ Hits         1605     1708     +103     
  Misses       1487     1487              
  Partials      132      132              

Impacted Files	Coverage Δ
internal/ingress/controller/kong.go	6.66% <0.00%> (ø)
...ingress/controller/parser/kongstate/credentials.go	83.50% <83.50%> (ø)
...l/ingress/controller/parser/kongstate/kongstate.go	24.70% <90.00%> (+4.20%)	⬆️
...al/ingress/controller/parser/kongstate/consumer.go	100.00% <100.00%> (+25.00%)	⬆️
...ernal/ingress/controller/parser/kongstate/types.go	60.00% <100.00%> (+60.00%)	⬆️

---

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update d2442de...6c28445. Read the comment docs.

[mflendrich]

Closing as WIP

[rainest]

Change that dismissed the old approval was just the retarget to next, correct? Re-approving since I don't think anything else changed.

To summarize some conversation from yesterday, there are two different properties I'd ideally like for the redacted dumps:

• You can diff dumps from before and after a change in K8S configuration to see what changed in the resulting Kong configuration.
• You can apply a dump to Kong via decK or a /config POST.

These unfortunately work against one another. Replacing sensitive values with just REDACTED prevents you from applying the config because it creates conflicting credentials and invalid certificates (we check key validity), whereas using random valid values breaks the ability to compare before/after dumps effectively.

I'd explored trying to generate garbage valid configuration that only changed if the underlying actual values changed, but that proved tricky to get right. Stripping certificates, credentials, and references to them from config can also produce valid configuration, but goes against @mflendrich's wish to not cherry-pick the redacted content.

In the interest of expediency, I think it makes sense to choose one or the other and live with the shortcomings. This version preserves diffability.

[mflendrich]

Thanks for the reviews!

Change that dismissed the old approval was just the retarget to next, correct?

Retarget to next + rebase.