get_job_logs with failed_only misses failed jobs after the first 30 jobs of a run

[jayhemnani9910]

Describe the bug

get_job_logs with failed_only: true lists the run's jobs with a single ListWorkflowJobs call and no per_page, so it only sees the API's default first page of 30 jobs. In a run with more jobs (a typical matrix build), a failure in job 31 or later is not found, and the tool answers "No failed jobs found in this workflow run". total_jobs in the result is also the page length (30), not the run's job count.

Affected version

v1.14.0 and main at 71ef826 (handleFailedJobLogs in pkg/github/actions.go).

Steps to reproduce the behavior

1. Run github-mcp-server stdio --read-only --toolsets=actions with a token.
2. Call get_job_logs with {"owner":"python","repo":"cpython","run_id":31008458461,"failed_only":true}. That run has 45 jobs; its 2 failed jobs ("Docs / Check for removed HTML IDs / Check for removed HTML IDs" and "All required checks pass") are both after the first 30.

Expected vs actual behavior

Expected: logs (or log URLs) for the 2 failed jobs, with total_jobs: 45.

Actual (v1.14.0):

{"failed_jobs":0,"message":"No failed jobs found in this workflow run","run_id":31008458461,"total_jobs":30}

Only the first page of jobs is requested.

Logs

N/A

I have a small fix with a regression test ready and can open a PR: list the jobs with per_page=100 and follow NextPage until the last page.