refine kill low-efficiency-job-alert email templates

[suiguoxin]

After this change, the template can also be applied for killing long-running jobs

[coveralls]

Coverage remained the same at 34.02% when pulling 4d17f2c on suiguoxin:alert-temp into fe18fd9 on microsoft:master.

[suiguoxin]

Test cases:

• manully send an alert by POST to https://xxx.openpai.org/alert-manager/api/v1/alerts with the following body & check email format:

[
    {
        "labels": {
            "alertname": "PAIJobGpuPercentLowerThan0_3For1h",
            "job_name": "test_user~test_job_0",
            "virtual_cluster": "default",
            "severity": "warn"
        },
        "generatorURL": "alert/script",
        "fingerprint": "6b8102e96c9e6b2a",
        "annotations": {
            "summary": "test_user~test_job_0 has a job gpu percent lower than 30% for 1 hour "
        }
    },
    {
    "labels": {
        "alertname": "PAIDebugJobRunningFor72h",
        "job_name": "test_user~test_job_1",
        "virtual_cluster": "debug",
        "severity": "warn"
    },
    "generatorURL": "alert/script",
    "fingerprint": "test_user~test_job_1 has been running for more than 72 hours in debug vc",
    "annotations": {
            "summary": "test_user~test_job_1 has been running for more than 72 hours in debug vc"
        }
    }
]