feat: [GCP Batch] Support passing standard machine types to the Google backend

[javiergaitan]

Context

This change rebases broadinstitute#7545 as well as broadinstitute#7518 on top of the deepgenomics/dg-ci branch. The goal is to port only the functionality necessary to support arbitrary machine types in wdl workflows into our fork of the cromwell repository.

Changes included

• WX-1810 WX-1830 n1/n2/n2d machine types, cpuPlatform on GCPBATCH broadinstitute/cromwell#7518 starts passing machineType as a separate parameter when creating an instance policy to be used when calling GCP Batch.
• GCP Batch: Support passing standard machine types to the Google backend broadinstitute/cromwell#7545 Creates a new standardMachineType runtime attribute in wdl workflows that allows us to configure such workflows to run on any predefined machine type on GCP.

Testing

1. Ran unit and integration tests

# Start sbt console
sbt
# Only relevant tests
testOnly cromwell.backend.google.batch.util.GcpBatchMachineConstraintsSpec
# All tests
test

2. Deployed and ran a wdl workflow to validate the change

version 1.0

task nvidia_smi {
    input {
        String docker_version
    }

    command <<<
        nvidia-smi

        touch .done
        echo "Finished at $(date)"
    >>>

    runtime {
        docker: <internal image>
        disks: "local-disk 50 SSD"
        memory: "32G"
        preemptible: 0
        gpuCount: 1
        gpuType: "nvidia-tesla-a100"
        standardMachineType: "a2-highgpu-1g"
    }

    output {
        File done = ".done"
    }
}

workflow nvidia_smi_wf {
    input {
        String docker_version
    }
    
    call nvidia_smi  as nvidia_smi_call {
        input:
            docker_version = docker_version
    }

    output {
        File done = ".done"
    }
}

3. Verified the machine type on GCP batch as A2 and the output of nvidia-smi showing the requested A100 accelerator.