Fix panic with Job resource diffing against an unreachable cluster

[rquitales]

Proposed changes

This PR ensures that we do not make a k8s API request during the provider's diff if there is an unreachable cluster. This currently occurs when the Pulumi program contains a Job resource with the replaceUnready annotation set to true. A panic would occur if we attempt to make the API call since our clients are nil.

Testing done:

1. Created a repro test case that fails with a panic (https://github.com/pulumi/pulumi-kubernetes/actions/runs/9228447658/job/25392833842?pr=3024)
2. Added logic to prevent the panic, and test passes subsequently without intervention (https://github.com/pulumi/pulumi-kubernetes/actions/runs/9228685506/job/25393667599?pr=3024)
3. Manual validation to ensure panic isn't trigerred.

Related issues (optional)

Fixes: #3022

[github-actions]

Does the PR have any schema changes?

Looking good! No breaking changes found.
No new resources/functions.

[codecov]

Codecov Report

Attention: Patch coverage is 0% with 5 lines in your changes are missing coverage. Please review.

Project coverage is 36.48%. Comparing base (b6f6eff) to head (2616e49).

Files	Patch %	Lines
provider/pkg/provider/provider.go	0.00%	5 Missing ⚠️

Additional details and impacted files

@@           Coverage Diff           @@
##           master    #3024   +/-   ##
=======================================
  Coverage   36.47%   36.48%           
=======================================
  Files          70       70           
  Lines        9160     9163    +3     
=======================================
+ Hits         3341     3343    +2     
- Misses       5489     5491    +2     
+ Partials      330      329    -1     

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

[EronWright]

LGTM

[blampe]

Makes me wonder if we should also check if the cluster is unreachable inside readLiveObject, seems like an easy mistake to make otherwise.

[rquitales]

Makes me wonder if we should also check if the cluster is unreachable inside readLiveObject, seems like an easy mistake to make otherwise.

Sounds good! Added an assert as a precheck for readLiveObject in https://github.com/pulumi/pulumi-kubernetes/pull/3024/files#diff-418e6e7d93aeacebe7c22d5fb674b48bfa15c6a514eb3ae4744b86d447595d8eR2684.