String collections in trigger metadata may contain nulls, objects, arrays, numbers and strings after conversion

We noticed this in the `messageIdArray` field of a TriggerMetadata of a service bus message batch. String collections may actually contain numbers.

After a thorough search, the culprit seems to be in `azure-functions-nodejs-library/src/converters/fromRpcTypedData.ts`

In the `fromRpcTypedData` function, instead of simply returning `data.collectionString.string`, `data.collectionString.string.map(tryJsonParse)` is returned:

```typescript
} else if (data.collectionString && isDefined(data.collectionString.string)) {
return data.collectionString.string.map(tryJsonParse);
```

What does `tryJsonParse` do?

```typescript
function tryJsonParse(data: string): unknown {
try {
return JSON.parse(data);
} catch {
return data;
}
}
```

So if a string ID is json-like, the json-like value is parsed and returned. Consequently, strings like `"123"` that are JSON-like will be converted as though they were JSON. In our case, IDs sometimes look like integers.

To add to the injury, if the string is JSON-object-like, the code will convert the keys to camel case. So if my message ID is the string `'{"BafflingConversion": true}'`, `fromRpcTypedData` will convert it to the object `{bafflingConversion: true}`.

This issue was introduced in efa5fb36. There is of course no obvious, sane and logical reason that the entries of a string collection should ever be considered as or converted to JSON, and whatever insane circumstance prompted the change isn't clear from the commit message.

@ejizba can you add some context as to why string collections should ever be considered as JSON? What was the intended purpose of the change?