fix: Make render_services exit non-zero on materialization failures

[mwarkentin]

Summary

• When a service fails to render (e.g. jinja2.exceptions.UndefinedError), the process now collects all errors, continues processing remaining services, prints a clear summary, and exits with code 1
• Previously the --multithreaded path would crash on the first failing future (hiding subsequent failures), and the non-multithreaded path had no error handling at all
• Full tracebacks are available via --debug and include customer/cluster/service context

Example output

Service FAILED: customer1 : cluster-a : getsentry — UndefinedError: 'dict object' has no attribute 'some_var'
Service unchanged: customer2 : cluster-a : common
...

4 service(s) failed to render:
  - customer1 : cluster-a : getsentry — 'dict object' has no attribute 'some_var'
  - customer1 : cluster-b : getsentry — 'dict object' has no attribute 'some_var'
  - customer2 : cluster-a : getsentry — 'dict object' has no attribute 'some_var'
  - customer2 : cluster-b : getsentry — 'dict object' has no attribute 'some_var'
make: *** [materialize-k8s] Error 1

Test plan

• Reproduced locally with make materialize-k8s — 4 failing services now show clear error lines and the process exits non-zero
• CI should now fail on render errors instead of reporting success

Made with Cursor