[Eng]: New-AzPolicyDefinition converts string value in policyRule to date, in a format that is not compatible with Azure Machine Learning Scheduler

### Description

When we are passing a json file which has a date as string in yyyy-MM-ddTHH:mm:ss format in policyRule, the New-AzPolicyDefinition is converting it to date object with yyyy-MM-ddTHH:mm:ss**Z** format. This policy is supposed to append a shutdown scheduler to Azure machine learning VMs. The scheduler does not recognize this new format.

Expected behavior: Leave the content of the json file as it is, without parsing the string as date.

Here is example of PolicyRule, where startDate is being converted to date with 'Z' at the end:

"policyRule": {
"if": {
"allOf": [
{
"field": "Microsoft.MachineLearningServices/workspaces/computes/computeType",
"equals": "ComputeInstance"
},
{
"field": "Microsoft.MachineLearningServices/workspaces/computes/schedules",
"exists": "false"
}
]
},
"then": {
"effect": "append",
"details": [
{
"field": "Microsoft.MachineLearningServices/workspaces/computes/schedules",
"value": {
"computeStartStop": [
{
"triggerType": "Cron",
"cron": {
**"startTime": "2022-05-02T00:00:00",**
"timeZone": "UTC",
"expression": "[parameters('cronExpression')]"
},
"action": "Stop",
"status": "Enabled"
}
]
}
}
]
}
}

Start by tracing New-AzPolicyDefinition's handling of the policyRule JSON, using the provided startDate example to reproduce the conversion. Identify where the string gains the trailing Z and verify that the completed change preserves the input date string unchanged for Azure Machine Learning scheduler policies.

Written by the indexing model from the issue text.