@confluentinc/schemaregistry performance optimisations: sync serialise function without fetching of new shema

Instead of fetching schema in serialise and using async function for it, it's much better to keep that function sync and fetch schemas separately.

If that is not possible, having a separate sync serialise function with separate function for prefetching schemas will be great.

What we have right now:
```
serialize(topic: string, msg: any): Promise;
```

What I am proposing:
```
serialize(topic: string, msg: any): Buffer;

fetchSchemaFor(topic: string); // even if this method is not provided, we can simply use original method for prefetching
```

Start by tracing the @confluentinc/schemaregistry serialize API and its schema-fetching path. Compare the current Promise contract with the proposed synchronous serialize and separate prefetching approach; done means the desired performance and API behavior are defined and covered by the relevant implementation tests.

Written by the indexing model from the issue text.