Improve error messaging when DCR is unsupported for remote MCP servers

[4t8dd]

Add clear, actionable error messages when OAuth providers do not support Dynamic Client Registration (DCR), helping users understand why authentication fails and how to configure client credentials manually.

Changes:

• Detect missing registration_endpoint in OIDC discovery document
• Recognize HTTP status codes indicating DCR is unavailable (404, 405, 501)
• Provide actionable guidance directing users to --remote-auth-client-id and --remote-auth-client-secret flags
• Add comprehensive test coverage for all DCR unsupported scenarios

Error message improvements:

• Before: "dynamic client registration failed with status 404: {...}"
• After: "this provider does not support Dynamic Client Registration (DCR) - HTTP 404. Please configure OAuth client credentials using --remote-auth-client-id and --remote-auth-client-secret flags, or register a client manually with the provider"

Fixes #2680

[ChrisJBurns]

Thanks @4t8dd , approved, let's see what the CI gods say 👍

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 55.61%. Comparing base (6800c99) to head (93d70dc).
⚠️ Report is 25 commits behind head on main.

Additional details and impacted files

@@            Coverage Diff             @@
##             main    #2696      +/-   ##
==========================================
+ Coverage   55.59%   55.61%   +0.01%     
==========================================
  Files         314      314              
  Lines       30445    30456      +11     
==========================================
+ Hits        16925    16937      +12     
+ Misses      12033    12032       -1     
  Partials     1487     1487              

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

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.

[ChrisJBurns]

@4t8dd There's some linting issues it seems, to see these locally run task lint (if there are some formatting issues you can run task lint-fix. Don't worry about the PR Size Labeler, that's a new action we've added and we're still ironing out the kinks to it.

[4t8dd]

@4t8dd There's some linting issues it seems, to see these locally run task lint (if there are some formatting issues you can run task lint-fix. Don't worry about the PR Size Labeler, that's a new action we've added and we're still ironing out the kinks to it.

sure. Thanks for the info. I will fix this then.

[4t8dd]

@ChrisJBurns I just updated this branch with lint fix. I guess the CI should be manually activated again.
One more issue is: I think CI should run against every PR to filter any lint issues before merging. Then it would be OK to review and so on.

[4t8dd]

I will check this again.

[4t8dd]

@ChrisJBurns I check this error

[FAILED] in [It] - /home/runner/work/toolhive/toolhive/test/e2e/thv-operator/virtualmcp/virtualmcp_inline_auth_test.go:208 @ 11/25/25 15:07:44.564
  STEP: Cleaning up test resources @ 11/25/25 15:07:44.564
• [FAILED] [30.062 seconds]
VirtualMCPServer Inline Auth with Local Incoming when using local incoming with inline outgoing auth [It] should proxy tool calls with inline auth configuration
/home/runner/work/toolhive/toolhive/test/e2e/thv-operator/virtualmcp/virtualmcp_inline_auth_test.go:162

  [FAILED] Tool call should succeed with inline auth
  Unexpected error:
      <*transport.Error | 0xc000471870>: 
      transport error: failed to send request: failed to send request: Post "http://localhost:30020/mcp": EOF
      {
          Err: <*fmt.wrapError | 0xc00012fa00>{
              msg: "failed to send request: failed to send request: Post \"http://localhost:30020/mcp\": EOF",
              err: <*fmt.wrapError | 0xc00012f9e0>{
                  msg: "failed to send request: Post \"http://localhost:30020/mcp\": EOF",
                  err: <*url.Error | 0xc000566330>{
                      Op: "Post",
                      URL: "http://localhost:30020/mcp",
                      Err: <*errors.errorString | 0x25b5530>{s: "EOF"},
                  },
              },
          },
      }

it looks have nothing to do with my patch.
I checked out the fix to main and run the test locally and it works good.
may I create another PR and close this one and check my changes?
let's just merge only my changes to see. Is it right for you?

[ChrisJBurns]

@4t8dd Yeah you're right, the e2e test are a bit flaky at the moment. We've raised #2741 to resolve this. I think you're changes should be good to do when the underlying flakiness has been resolved 🚀

[4t8dd]

I close this and open a new one based on the latest update.