Remove unnecessary JSON parsing in BackupJob

The `BackupData` project has a redundant JSON parsing step when processing MongoDB documents:

```
BsonDocument → ToString() → JSON string → Parse into JObject → Extract fields → ToString() → Upload
```

The `JObject` is only used to extract two fields (`_id` and `_meta.updated`). We can read these directly from the `BsonDocument` and skip the parsing entirely:

```csharp
// Before
var jObject = JsonConvert.DeserializeObject(document.ToString()) as JObject;
var blobName = jObject["_id"]?.ToString();
var updated = jObject["_meta"]?["updated"]?.ToString();
var json = jObject.ToString();

// After
var blobName = document["_id"].AsString;
var updated = document["_meta"]["updated"].ToUniversalTime();
var json = document.ToJson();
```

This removes the Newtonsoft.Json dependency from the production build and eliminates a redundant parse/serialize cycle. The output format stays the same (JSON files in blob storage), so there's no breaking change for consumers.