Add Google Batch job runner

[ksuderman]

Summary

Adds a new job runner for Google Cloud Batch, enabling Galaxy to execute jobs on Google Cloud's managed batch computing service.

Features:

• Container-based execution with automatic tool container discovery
• NFS volume mounting for shared data access
• CVMFS support for reference data
• Dynamic CPU/memory allocation from job requirements
• Job monitoring, recovery, and cancellation

Configuration

Add the runner to job_conf.yml:

runners:
  gcp_batch:
    load: galaxy.jobs.runners.gcp_batch:GoogleCloudBatchJobRunner
    workers: 4
    project_id: my-gcp-project
    region: us-central1
    nfs_server: 10.0.0.2
    nfs_path: /galaxy/server/database
    nfs_mount_path: /mnt/nfs
    container_image: quay.io/galaxyproject/galaxy-min:25.1

destinations:
  gcp_batch:
    runner: gcp_batch
    params:
      machine_type: n2-standard-4
      requests_cpu: "2"
      requests_memory: "4Gi"

Required GCP setup:

1. Enable the Batch API (batch.googleapis.com)
2. Create a service account with Batch Admin and Compute Admin roles
3. Configure NFS (Filestore) accessible from Batch VMs on the same VPC
4. Authenticate via GOOGLE_APPLICATION_CREDENTIALS or service_account_file parameter

Authentication: The service_account_email parameter is optional and primarily useful for development/testing. In production, use a workload identity pool for secure, keyless authentication.

Key parameters:

Parameter	Description	Default
project_id	GCP project ID	Auto-detected
region	GCP region	us-central1
machine_type	VM machine type	n2-standard-4
container_image	Default container	quay.io/galaxyproject/galaxy-min:25.1
nfs_server	NFS server IP	Required
requests_cpu / requests_memory	Resource requests	1 / 2Gi

Test plan

• Unit tests pass: pytest test/unit/app/jobs/test_gcp_batch_runner.py
• Manual testing with GCP Batch infrastructure

License

• I agree to license these and all my past contributions to the core galaxy codebase under the MIT license.

[nuwang]

Just some preliminary comments.

[ksuderman]

I have been asked why not use the new Pulsar enhancements introduced in #20862

1. We did not develop this in isolation and we did consult with @jmchilton before starting work on this
2. The GCP Batch job runner is optimized towards a deployment on AnVIL, or at least Galaxy instances deployed to GCP.
3. The two approaches to dispatching jobs to Google Batch are not mutually exclusive. I had Claude write a short summary of the pros and cons of each approach. TLDR: it depends, and if Galaxy can be configured to use Google buckets for its object store any differences in I/O are negligible.
4. There are some bugs and features missing from the Pulsar GCP runner that means we can't use it on AnVIL in its current state. I will PR fixes and updates, but this runner is tested and working now.

[jmchilton]

Can you fix the remaining linting issues?